This is the seventh part of The Ouroboros Sequence, a series on puzzle games.
She believes she has exhausted all of the beautiful possibilities in the puzzle rules before her. But there are only 15 levels. 15 exquisite, perfect creations, full of nuance, each offering something unique and precious. It’s too short. Jonathan Blow didn’t make billions with 15 line puzzles. How many puzzles did The Witness contain in the end? 500? 600?
“There’s only one thing for it,” she muses. “Make bigger puzzles.”
The technical term for this design transition is jumping the shark.
Suppose we set ourselves a challenge. Make a block pusher… but we’re only allowed to work on a 2×2 grid. This pocket reality is too small to support puzzle life. We need to stretch our canvas if we want interesting things to happen. Every time we enlarge the universe, new puzzle opportunities spring into life, scenarios that were not possible before. On a 3×3 grid, the classic three-block-wall problem emerges, as depicted in the following image, where the player must somehow get to the other side.

Still, we cannot keep inflating the universe forever. If we present a player with a 100×100 Sokoban puzzle, they would have every right to pop round and kick our collective asses. There’s a reason few are willing to entertain the ridiculous Taikyoku Shogi.

More moving parts means more mental juggling. Faced with an enormous possibility space of outcomes, players are more likely to default to brute force instead of reason. Also, it’s unlikely you’ll find truly fresh challenges out there; they’re just the same kind of puzzles with an injection of the complexity steroid. Do not make the same mistake millions of wannabe game designers have fallen for – puzzle complexity is not the same thing as puzzle joy. You know how I feel about lots of numbers. ‘Complexity’ and ‘Interesting’ are two different words. I know this. I checked them out. The dictionary was on my side.
I brought PuzzleScript title The Flames (Rosden Shadow, 2017) to your attention last time because its ten levels are small and tight. They feature just a handful of elements – pushable wood, wood wall, stone wall, fire barrel – which are easy to understand. Yet the levels are fiendish.

This is precisely why I like the light deflection-refraction puzzler Archaica: The Path of Light (Two Mammoths, 2017). The largest puzzles never become Satan’s chessboard. So few moving parts, yet so much grief. There is nothing quite like the austere beauty of a compact puzzle.

Compare with Operator Overload (Benn Powell, 2017) which has a lot more puzzle content, but often needs to dig deep for that content. And by deep I mean large; see exhibit A below.

Large puzzles can have a fatiguing effect on the player, particularly if they’re feeling in a more casual mood. A glance at a large grid filled with identical elements and, as the Klingons say, today is a good day to sigh.
Time to be a little more charitable. There is no neat, easy distinction between interesting and uninteresting complexity. Different puzzles operate in different ways. I would wager a 20×20 grid with just four pieces to play can still be interesting.
Dissembler (Ian MacLarty, 2018) proves that visual size is not a reliable guide for complexity. Dissembler is a recent favourite of mine that I still play daily, where you must flip tiles to make matches of three and eventually clear the board. It starts simple but gradually ramps up. The levels are never huge but because every single tile must be removed, it means every single tile is involved in the solution.
But Dissembler also throws in the “tile-within-a-tile” which exposes a different colour upon match. Physical size is therefore deceptive as the number of matches required may be larger than you expect. For a game like this, move count is a better way of measuring complexity in turn-based puzzles. The solution to the following puzzle is 11 steps long which, by Dissembler standards, is pretty high consisting the dimensions.

I wrote extensively on The Talos Principle (Croteam, 2014) a few years back. It’s pretty good at keeping the puzzles tight, but the optional stars often require brilliant masterplans to be conceived and implemented using components from multiple puzzles. This is the perfect kind of large challenge. Optional, inventive… seductive. The excellent Road to Gehenna DLC does inevitably tread down the large level path a few times including one the designers admit is large: they call it ‘Goliath’.

But Goliath demonstrates what makes a big level work: the possibility of breaking the puzzle down into something manageable. Analysis of Goliath, investigating cause and effect, will gradually clear away the fog of complexity that obscures the solution. It really isn’t as hard as it looks. This is why Dissembler, mostly, works even for the more complex challenges because Dissembler is ultimately about travel: a good player will know where colours need to be and how far it is possible to move them over the duration of a puzzle.
We all have our personal limits, though. I have a mental breakdown when it comes to cog-based puzzles: they become overwhelming with just a handful of cogs. This explains my allergic reaction to These Robotic Hearts of Mine (Alan Hazelden, 2011) when I dabbled with it at the Eurogamer Expo. I won’t touch Prismatica (Loomus Games, 2015) with a barge pole. I just can’t see through these types of puzzle.

I found the genius Cosmic Express (Hazelden & Davis, 2017) also belonged firmly in the Dissembler camp of “manageable complexity”. In each Cosmic Express puzzle, you need to draw a rail that lets a train pick up every passenger and drop them at the stations they need, subject to various constraints. Even the largest puzzles successfully focus your attention on issues or tasks, such as different ways to shape the route or reorder the action… until Cosmic Express jumps the shark with its final level, Nova 7.
I was already experiencing size fatigue when I reached Nova 7 but there was a moment of pure horror when it was revealed to me.

A vast blank canvas. It was clear what kind of level this was going to be. You’d always be short of space and never be quite sure what you were doing wrong. It wasn’t about solving a puzzle but approaching a giant project of optimisation. In fact, this is precisely what The Flames is, because solving levels requires the player to minimise the number of steps taken, but developer Rosden Shadow was mindful to keep things small which keeps you focused on the how. I hated Nova 7. Although I had a personal feud with a level called Andromeda 14, it was Nova 7 that felt like a betrayal. And if you want to take Cosmic Express to its ultimate conclusion, you have to solve Nova 7 more than once.
Of course, there are designers who play with size deliberately for effect. In Stephen’s Sausage Roll (Increpare Games, 2016), the player is thrown into ‘The Great Tower’ without any warning to brace for impact, which looks like this:

The discontinuity between this level and everything that has gone before is so shocking that the player is under no illusions that this is “a designer accident”. That shock is deliberate. If this was a horror game, we’d call it a jump scare.
But it suffers from exactly the same problem as all large levels: the sigh. I am most likely to abandon a puzzle game when the levels get large. The more time I spend on a level, the more my love and interest ebbs away. I make that promise of “I’ll come back when I have some time”. That’s like promising someone you break up with, “I hope we can still be friends”. Good luck with that.
Incidentally, the last level I played of Stephen’s Sausage Roll was The Great Tower. But don’t worry. I’ll come back to it when I have some time.
Appendix: Level 25
In recent years I’ve become unhappy with the last “real” level of The Citadel (Joel Goodwin, 1991). It is a classic BIG FINAL LEVEL.

It’s not as bad as it might seem. If the player has become experienced with the game’s mechanics, this is the kind of thinking process I’d expect to occur:
- The exit is behind square pits. You need the block.
- The exit is also behind a sea of round pits. What is the minimum number of boulders needed to reach the exit? 6.
- How many boulders are needed to get the block out? Moving the block anywhere requires you to move space in front of and behind it. If you want to move the block directly upwards into the shortest exit row, it will cost you 13 boulders – and you will not be able utilise boulders trapped against the top row of the puzzle. In contrast, the more seemingly wasteful horizontal extraction approach can be done in 12 plus you can use boulders trapped on the northern edge of the puzzle. This is the bit where players are more likely to dive in and make a mistake on first pass, not realising how much space needs to be developed around the block for extraction.
- Right, you must find 18 boulders. The player should decompose the left half of the level in the following manner.
- First, there’s a 2×2 block of boulders, they are just noise, a wall. There are 4 easily-rescued boulders around this block, but the fifth is tied into the northern boulder cluster. We can come back to that.
- Next, the southwestern boulder cluster. I can squeeze 4 out of this, including 2 boulders trapped against the wall, if I push in the boulder at the top and trap it against the wall.
- Finally, the troublesome northern cluster. Even now, I don’t have any heuristic to maximise the number of boulders, I just stare it and try to find something that doesn’t trap and lock boulders away. Sacrifice is vital to get the necessary ten boulders but where is this sacrifice committed? The correct sacrifice is to push boulder A up then push boulder B all the way to the back; a single boulder sacrifice. Note that obtaining boulder C is dependent on a clean solution to southwestern boulder cluster.
But have players learnt everything they need to by this point? Will doubt undermine their rationalisations? Will they plan or just dive in? Is this level enjoyable to solve?
Download my FREE eBook on the collapse of indie game prices an accessible and comprehensive explanation of what has happened to the market.
Sign up for the monthly Electron Dance Newsletter and follow on Twitter!
I kind of hated that level of The Citadel, but it felt good to solve largely because I didn’t think in the order you described.
I’d already got out (what turned out to be) the maximum number of boulders available, having figured out that pushing the block across the top was the way to go – but I fell one short, not having realized there was a better way to get the block clear of the field of holes.
So the last step of my solution was figuring out the better way to extract the one block (hinging on the fact that there is ground underneath where the block starts off) – which was a much more satisfying finish than counting boulders!
Here you seem to conflate too-big and too-complex, but I want to argue that too-big is a problem even if a level isn’t too-complex. My first piece of evidence: The Citadel. I started playing the PuzzleScript version and couldn’t stand it! I think I stopped on level 3 or so, it was a block-pushing room where you had to take a convoluted route from one side of the room to the other, multiple times. A good puzzle is almost always as small as it can be, with no unnecessary walking.
At some point I’ll write up the history of Nova-7. I’m not expecting to convert anyone to my side who hates it, but I think I can justify most of the decisions that made it how it is.
Phlebas
As I was writing the appendix, I realised this was way too technical. No one thinks like this. So it’s an approximation of the thought process. More likely people were using hunches to make their way through it; that’s probably how I did it originally. The original design of Level 25 was just random: I made this giant set of pits then threw in lots of boulders and hoped it was possible. I think it was edited a little, but not that much.
The technical analysis I put forward here reveals how subtle the nature of the solution is: being able to rescue boulders from the edge makes it possible. I had always assumed there was probably multiple solutions but I think there isn’t?
But, yeah, this kind of large level just feels wearying. It’s clever and you feel good if you solve it – but I don’t think it’s engaging.
Alan
Ha ha, oh my God I hate that level of The Citadel. The big W, right? Every time I was testing I wanted to scream when I encountered it. I think I said as much on the commentary. Whyyyyy. I left it in because it was supposed to be an historical reproduction. But I’d be quite happy to completely change the level…
Conflating too big and too complex? Probably. The focus is meant to be the psychological impact of a “large level” which is often tied to its complexity. If I was to be exact about it: largeness per se is not the problem but it is often correlated with over-complexity. As a result, a fairly simple level that doesn’t reveal its solution secrets to us quickly can be tarred as “boring large level”. I got stuck on The Great Tower and I’m having difficulty going back to it even though everyone tells me it’s not that bad.
However, something I have definitely conflated is “design advice” with “player experience” which the language has blurred. I wasn’t really offering design advice. This whole series is figuring out why puzzles go wrong – because I wanted to understand my Snakebird experience. (Unusually for me, this series is focused around negativity.)
What I said here about Nova 7 is the endpoint of my experience rather than the beginning, how I felt about it in the end. I flinched at the open space but assumed it would probably open up to me with careful analysis… and that never happened. I’ll be expanding on this later on, because Nova 7 drives two different essays later in the series (provisionally – A HOLE IN MY CHEST / ALL ROADS). Instead of Nova 7 turning up in three different episodes, I originally planned a single essay just on this level.
I remember you mentioned that Nova 7 wasn’t your first choice of “final level” but having all the different elements – portals, cross pieces, wildcard stations and all three aliens – made Nova 7 the more apparently suitable choice for final level.
Regardless, I absolutely adore Cosmic Express. I finished everything which is *rare*.
“I flinched at the open space but assumed it would probably open up to me with careful analysis… and that never happened.”
This is my experience too. I solved Nova Express the first way honestly and the second way with a hint (I looked at the solution, retained where the track goes at the beginning, and then got it without too much problem), but after the second solution especially I was like… I’m not sure what problem it is that I solved?
The lovely thing about a big level or a complex one for me is figuring out how it falls apart into separate problems. I’ve mentioned Blocks With Letters On once here, it’s a Sokobanish or possibly Snakebirdish game about getting the Blocks to a goal area where they spell a word (you can move the blocks themselves but they’re usually subject to gravity, though some aren’t and some levels have features that switch gravity on and off for a block). The last level of the second game (More Blocks With Letters On) is one of my favorite puzzle levels ever–after figuring out your target word (which requires a pretty basic analysis of what letter changes are possible) it’s a puzzle to figure out how you can possibly get anything out of the initial starting area, and then how you can do that without messing everything up, and then what you can do out of there, and then what the endgame is, and then what you had to do at the beginning to prepare yourself for the endgame… it seems like an unholy mess at first but by the end you’ve broken it down in a bunch of stages which is incredibly satisfying. Then the previous level (IIRC) is a bit more of a classic huge level in its way–you have to bring letters across a gigantic field of blocks that are turned on and off by a complex array of switches, and at first it looks like finding a bath across a bowl of oatmeal by jumping from lump of porridge to lump of porridge, but a little experimentation helps you find the paths (because when you hit a switch a posslbe path becomes manifest, like one of those numbers popping out of a colorblindness test).
OK, where was I going there… so Nova 7 doesn’t fall apart into separate tasks like that for me even after I’ve solved it, although there are certain microproblems that I was able to focus on (I had everything except what I needed to do in a certain corner, then I figured out what to do in the corner). The first time the way I solved it was by saying “Alan likes to make it so the obvious connection closes off the real solution, so I’m going to start it the way that seems to make the least sense without provably making the game unwinnable.” Refusing the Schmuck Bait. (You know looking back at old comments I’ve said basically all this, in the comment where I think I first mentioned Schmuck Bait as applied to puzzles.)
This was different than the Turn 1 Dick Move Steed was complaining about with Stephen’s Sausage Roll, where you could make the level unwinnable at the beginning–do the rest of the level and find you’re stuck. Every level has to be played twice, he complained, but that’s part of how the level starts to make sense. You do something, get to the end, find you’re stuck, and now you realize that one of your tasks in the level is not to get stuck that way, or to find a different way around things so you’re never stuck at all.
About the Citadel: I never got that level (partly because of save game issues) but I like the analysis of how it comes apart. But I fell down at the “how do you get the block out” stage. I didn’t realize I should bring it out to the west and was trying to make a ziggy-zag path to the exit which was going to require ridiculous amounts of boulders! The issue here for me is that experimenting with different ways of getting the block out was going to require solving most of the boulder problems, so it takes a lot of effort to work on the part I need to work on. Here’s where Steed’s complaint about SSR comes in–once you realize what the problem is you still have to remember and redo all the other stuff you’ve worked out, so there’s a lot of friction in your experimentation. Cosmic Express is good on this because you don’t have to keep a unified track at all times–if the middle of your solution is messed up you can retain the beginning and end without having to reconstruct it once the middle is solved.
Some more thoughts I have, about The Great Tower–there’s the bigness of the open space, which makes the level easier because it’s giving you room to maneuver, and the bigness of the number of sausages, and the real killer is the bigness of the number of new mechanisms that are introduced (which it’s getting you to experiment with) and the attendant fact that the work you need to do around the grills doesn’t conform to any of the patterns you’ve learned up to now. And about the huge everything-is-interconnected puzzles like Corrypt and Promesst and the dream content of Do You Want To Build A Snowman and why I admire them but don’t enjoy them, and it seems like Sean Barrett doesn’t enjoy them either oddly enough. And maybe how SSR does something similar but different. But I’m tired and should go to bed.
Hey Matt,
I was really hit by the Turn 1 Dick Move with Nova 7. The first time I walkthrough’ed it but the second time I was determined to defeat it. The route that made most sense to me was through the portal, otherwise you start gumming up the board with “unnecessary rails” is what I thought. And I tried for ages and ages. I then looked back at the previous Nova 7 solution and looked at the start – it didn’t use the portal. Once I stopped using the portal, that was it, I solved it instantly. (I’ll expand on this in later articles.)
But, like you, I don’t mind T1DM if it you can make sense of it and the level isn’t too laborious. All of these puzzle games are about sequences and it’s incredibly difficult to limit puzzle designers to puzzles which don’t go wrong on With Nova 7, also like you, I didn’t see that. I think admitting puzzles have beaten you is a difficult thing, but to go further and claim the puzzle doesn’t feel fair – I can tell you it took some gonads because I’m sure some people, Alan included, can understand that puzzle. I don’t have the stats but I suspect Nova 7 leaves many people floundering and even when they plugged in the correct route, they feel empty: they planted the rails and pressed the buttons but never engaged, never solved.
On the Citadel: You can download the HTML from the page, that might remember the save game better than playing inside itch.io. But on the “ziggy-zag path” – part of me is flabbergasted anyone would think of doing it that way 🙂 I am thinking of The Flames, where any directional changes are expensive to the point of failure, and Level 25 manifests this problem in the block extraction. But your concerns about labour and recall around larger puzzles are valid, particularly for this one… however, there’s one thing that stops me going 100% on the “having to remember” argument. I will get into this later (I know I keep saying this) but if your solution is dependent on not having to remember what you did and why there’s a possibility that your brain hasn’t understood the puzzle. Redoing it helps to cement understanding.
I think what is off-putting with The Great Tower is not just the perception of size but the fact you have to deal with eight sausages. I think that’s a record at this point? I discovered by accident (as everyone does) the brand new method of moving the sausages (trying not to spoil) but that didn’t seem to get me closer to a solution.
Anyway because everyone was moaning at me about it I solved The Great Tower last night.
on the “ziggy-zag path” – part of me is flabbergasted anyone would think of doing it that way
I have to run in to an appointment now so I only have time to inform you that it is ON.
Oh, I finished the post now! Congratulations! The game isn’t going to do you like that again… well, at least not for a little while.
My Twitter thread on The Great Tower.
There seems to be some words missing in the middle of my long comment above. Ah well.
With Dis Pontibus I attempted to walk the edge between size and complexity constantly; I limited the number of pieces and the possible layouts, but set free the generator to “mine” increasingly complex levels. I think the result is pretty acceptable for a procedural “puzzler” 🙂
“Frenemies of mine had spoken.” I feel named!
Y’know the thing about The Great Tower is that it is not in fact a great puzzle qua puzzle. By which I mean: If it were placed later in the game, after you’d mastered the mechanisms involved (however you would’ve done that), it wouldn’t be a particularly interesting level. Your solution feels like a mess because there are lots of different ways to do it, and many of them are messy. You wouldn’t believe how ridiculous my first solution was–univat tbg frireny fnhfntrf va cynpr, V unq bar zber gb trg bagb gur hccre yriry, naq V chfurq vg bire gb gur onpx jnyy naq gura fubirq vg bire gb gur fvqr naq xabpxrq vg nybat gur fvqr jnyy hagvy vg jnf jurer V arrqrq vg gb or. The real thing about it for me is the endgame, what you do with the sausages when they’re where you need them to go. The gigantic tower and the wide open space around it give you a combinatorial explosion of ways to get there.
That’s part of what makes it so off-putting–the previous levels were so tight and confined that you only had a few possible ways to irreversibly change the state at the beginning, and in TGT there are tons. Tons! You don’t face the Turn 1 Dick Move per se (I don’t think it’s possible to make the level unwinnable the first time you strike a sausage) but precisely for that reason it’s intimidating–when it takes you several turns to screw yourself over it’s harder to see how you’ve screwed yourself over.
The tower sits there like an upraised middle finger, saying “What are you going to do with me now?” but that’s a bit of a first impression, you’re going to hit something similar-looking soon which not to be spoily intimidated the hell out of me at first but when I got into it was not at all one of the nastiest levels of that bunch IMO. The big thing is that it’s sitting there in empty space. And since the endgame is where you really can burn yourself, but it takes a lot of work to get to the endgame, you get the problem I had with your Citadel level–the cost of experimenting is high because there’s so much work to prepare for the experiment.
What made it tolerable for me was what Pippa Warr talked about in her review: “I started tinkering with it and worked out how to cook a single sausage. From there I could cook four sausages. Then I worked out how to cook a sausage on the higher level of grill pads. Now I’m alternating between fiddling and contemplating. I’ll still have that knee-jerk dauntedness if I leave the level and open it again but I know I’ve made progress and that the manageable chunks will eventually come together into a solution.” There are these little things you can do even if you know you’ve broken down the tower wrong, and when you really really really work out the higher-level cook it gives you a goal for how to break the tower down. So in the end I can remember what I did.
The reason there’s a not-great-once-you-understand-the-mechanics puzzle here is because it’s there to teach you the mechanics. And maybe the idea is that it has to be huge enough and frustrating enough to force you through a lot of experimentation so you’ll have really mastered the mechanics? I’m not sure how necessary it is, though I can talk about How SSR Introduces Each Of Its Mechanics some post-spoliation time. To some extent I want to say that the mechanics are new and unexposed even though they make a sort of physical sense–there’s nothing in the first set of levels from a mechanical standpoint that guarantees that you’ll be able to stick a sausage by moving into it when it’s immobilized, but of course when you do it makes sense, because you can see that your fork is pointy. And of course the game doesn’t explicitly tell you anything about the basic bumping, rolling, and swinging mechanics, you figure that out from the beginning. The only explicit instructions are the “use arrow keys to move” thing written on the sign that you can only read once you’ve moved to it, using the arrow keys. So in this way all the new mechanics are playing fair. (Though I think I might have had to look at a walkthrough to figure out how you climb ladders.)
Incidentally SSR has the reputation of not doing the “Here’s a trick to solve this puzzle, you’ll be using that trick in elaborated ways in future levels” (the three-block problem is like that), but it does to some extent. Like the level you screenshotted in one of the earlier posts, “Infant’s Break,” is really all about figuring out what you can do with grills in patterns like that–and you do wind up using that a lot, it’s just that it’s almost automatic at the end of some impossible level. That’s one of three levels in a cluster right by the beginning, two of which are some of the less difficult levels that teach you important tricks and the third of which is Lachrymose Head.
BTW I didn’t have a T1DM issue with Nova 7 so much, I hit on the use of the portal pretty early on as a hypothesis… it was really a combinatorial space issue. Not knowing when I was beaten.
I have things to say about Promesst and Corrypt still, and what they do to the possibility space, but I must go to bed. Free Ice Cream Day was brutal.
Marcos,
I fiiiinally tried Dis Pontibus last night. Interesting how some of the puzzles are obnoxious with such a small number of components! I can’t think offhand of any puzzles that it resembles – were there any other titles that inspired it?
Matt,
When I discovered I could *pneel fnhfntrf* that set me off briefly on another wild goose chase in The Great Tower. However, I think that definitely is useful for later levels and considering this additional mechanic for use. But I haven’t completed any of the new puzzles yet!
With large puzzles perhaps there should be a way of “saving states” rather than just restart/undo. Oh, by the way, absolutely excellent that “restart” is recorded as an UNDOable action! Life. *Saver*.
But I also think The Great Tower is one of those puzzles where reputation carries you though. In the absence of praise or knowledge of Stephen Lavelle, I’m pretty sure there’d be even more walkouts. Like Starseed Pilgrim. People only persisted because others said it was good to persist. It’s increpare! There’s a reasonable solution here somewhere!
Something I don’t really don’t like about The Great Tower is that 3D legibility is important and this is something SSR does not do well. You’ve got this zhfu bs fnhfntrf cvyrq hc naq orvat gb ernq UBJ gurl’er cvyrq vf pevgvpny. Gung’f bar bs gur ernfbaf V nffhzrq vg jnfa’g vzcbegnag. Some of the mechanics feel like a bit weird, too, like how to climb ladders as you stated or what you have to solve The Great Tower.
I was going to dip into the “giganto” puzzles like Corrypt, Promesst, Snowman’s dream… but they’re a special kind of puzzle. I mean, you know what you’re getting into – even if they have the same “holy crap” problem.
What I don’t know about is free ice cream day.
Free Ice Cream day is a day when you, through poor planning, wind up having to walk straight from work to the children’s school, and one of your children says on seeing you “I want ice cream,” and you say “It’s good to want things” (because you are an overly snarky parent), and that child’s friend asks if you’re going to Free Cone Day and that child’s friend’s mother says “I’m so so sorry,” because Free Cone Day is a day when your town’s iconic ice cream store is giving away free ice cream cones and since people like ice cream as you can imagine the line is LONG, but you decide that the children will enjoy it so you go, and the line is indeed LONG, and there are Adventures, but they (and you) do enjoy the ice cream, yet nevertheless it is enough to leave you wiped out through the end of the next day’s teaching.
Thanks for playing it! The main inspiration for the format (islands with difficult puzzles) was, of course, SSR.
The mechanics can be considered a mixture of games like Klotsky and some take-apart physical puzzles, I suppose.
At first the objective was to make two special pieces touch, but I later changed it to “reach to the other side” in order to add exploration to the puzzles more naturally.
About SSR: You’re totally right about the camera/3D perspective being a pain. There was at least one later puzzle where I was stuck for a long time in part because I couldn’t tell how big a gap was… though after I figured that out I was still stuck for a long time. I had similar issues with English Country Tune sometimes. I even abandoned it early on for a little while on a level where I couldn’t parse the 3D. The movement scheme of ECT makes it necessary to take away your camera control (you can go on the sides and bottom of the levels and the camera moves with you) which doesn’t seem as true in SSR.
The other stuff you say about the mechanics being a bit weird maybe does explain why the game needs a level like The Great Tower. There are other mechanics that aren’t weird but aren’t obligatory, like the sausage-spearing I mentioned, or the fact that you can push a sausage with a sausage. That seems (fairly) natural because of the way the sausages roll, but it doesn’t have to be that way, and it isn’t that way in most Sokobans. (In The Citadel it’s more natural that the round boulders jam and the square blocks can be slid in rows, because the lack of animation suggests sliding rather than rolling.) There are some other mechanics that come naturally, and some where you have to think of something that sort of makes sense with the intuitive physics, and at least one where the level design smacks you in the face with it (I’ve talked about this in rot13 elsewhere in the comments), but here there are really a lot of nonobvious ones (and some whose rules I don’t think I can entirely formalize after playing through the game about four times), and maybe the only way to get you to learn them without an explicit tutorial is by throwing a huge mess like this at you.
About the huge things like Corrypt, Promesst, and Snowman’s Dream; in the first two there’s this moment which I think of as the game coming alive and firing back at you. As Sean says in the piece I linked, “the later stage of the game introduces some chaos–you’re simultaneously solving puzzles and creating new ones for yourself due to the non-local consequences of your actions.” This blows up the possibility space immensely and also creates a lot of delayed-action walking deads where you can’t be sure exactly why you’re stuck and how far back you need to go. There’s sort of a spectrum here–in Corrypt the changes you make are really non-local, in Promesst less so except that the things you can do in the first stage are also often non-local, and in Snowman everything you do is fairly local except that one you can do in one place may affect what you can do in the next, and in the next, snowballing all the way across the board.
This makes it really hard to break things down into manageable chunks. Maybe impossible! I’ve never solved them–in both Corrypt and Promesst I saw the twist, said “Wow that’s cool,” did two things, and stopped playing. Besides the manageable chunks thing, with effects that go across the entire map it’s physically impossible to visualize what you’re working with unless you write something down and take notes on it. And I’m not someone who likes to take notes on puzzles. The loop between experimenting and figuring out whether the experiment works is so huge it eats the whole game. Also and related, none of these have entirely satisfactory undoing mechanism–when I undo in Promesst/Corrypt I don’t even know what I’ve undone.
SSR also has a moment where the game starts firing back at you, which I daren’t say anything more about, except even then there are still levels so you at least know what problem you’re trying to solve.