It's a fine level, working in plenty of variety in game elements. Really, the issue is that by making use of all that variety, that's more opportunities to be susceptible to perceived flaws that go all the way back to the burden of the people who originally programmed the source games
and how they defined their elements should work, before they were picked up and faithfully implemented in RnD.
One of the concepts that comes up a lot in games of this genre, at the micromanaging level, is something called "reading order", which dictates what order things are set off in after every opportunity for the player's input to be reflected. For example, suppose you have this 3x3 snippet of a level:
Each of the tiles 1 through 9 in turn will be checked for interactions, on every frame. So, let's suppose there's a gravity-prone element in square 2, another gravity-prone element falling out of 5 into the vacant square 8, and a growing wall in square 6. Perform the checks in order:
- 1 is irrelevant and doesn't do anything.
- 2 can't fall because 5 isn't empty yet, so it does nothing.
- 3 and 4 likewise have no interactions.
- 5 finishes vacating its square,
- Now we get to 6. If this is one of the frames on which it can trigger expansion, it will go "Dibs on square 5!" and immediately close up that square, sending out another growing wall in its formative stage before 2 gets another chance to look at it. If this is a dormant frame for 6, then it won't do anything right away, and on the next frame, 2 will get to look at the empty square and start falling first.
Interestingly, if you flip that configuration around so the growing wall is in 4 rather than 6, gravity will win 100% of the time: after any action taken by 5, 2 will always get the next look at that square on the following frame, before 4 does. This means that if you could rearrange the rooms so that one has its entrance/exit on the right side rather than the left, the issue of the gems sometimes being split would be eliminated. But the fact remains that this is a reasonable way to want to set up a room, and having the rules work so that the mechanism only works one way around and not the other, at least not without risk of malfunctioning, is a case where sensible level design is at the mercy of the rules. If growing walls were programmed with enough state knowledge so that they wouldn't grow into a square until they observed it empty for two consecutive frames, that would be a sufficient change that they would always defer to gravity elements in a fight over a newly vacated square, no matter which side the growing wall was positioned on. Then again, that change would also break the last levels of both tutorial sets, which want you to keep jumping up and down over a growing wall so it has room to grow under your feet, enabling you to climb against gravity. Different rules have different consequences.
Playing in slow motion enables you to have a watchful enough eye to analyze a lot of things like this closely. Here's another example.
This is the room immediately before the first Supaplex room, designed to test landmines, and DX-Boulderdash-style traps. Mines, of course, are always deadly, where the point is to notice the visual bulge that prevents them from blending in seamlessly with the surrounding dirt. Fair enough. Traps, meanwhile, cycle between safe and deadly phases at irregular intervals, where the safe state is visually distinct from dirt only by a small black dot in the center, which animates into a large sinkhole during the deadly phase. However, for the first three frames and last three frames of the deadly phase
, its animation shows a graphic that is completely identical to that of the safe phase. Unless the DX programmers intended that the presence of their new trap tile in a level should have induced all players to proceed with extreme caution, only ever dispatching of the traps by snapping and only one square at a time, it's inevitable that someone eventually walked up to a trap that, by every visual indication, was in its dormant state at the time, attempted to clear it out, and was greeted by the equivalent of "Whoa there! You couldn't have possibly seen anything to detect it, but this tile flipped to the deadly phase some time in the last three frames before you attempted to walk into it, so congratulations! You're dead now!" Granted, the chances of this are only three frames out of the length of each safe cycle, far less than the 50% chance of a right-side growing wall misbehaving against a gravity element that wants to fall into the same square. But it's a separate, cumulative chance for every instance of the DX trap that the player is forced to walk through for lack of another path that goes around it, and those chances add up, again unless they take the ultra-cautious approach of only snapping to get rid of the traps.
For what it's worth, the Supaplex equivalent of the trap (buggy wall) errs in the opposite direction. There's a period of time where it's just started visibly sparking up, but hasn't become deadly yet, and you can safely enter the square to put it out for good. This at least insulates their programmers from complaints on this charge, as one less thing that can reasonably be considered "unfair death".
RnD, for its part, does mitigate the potential for perceived injustices like that to irrevocably throw away large amounts of progress into a level, by virtue of its tape system. Even if you don't micromanage it to play in slow motion, anyone can still replay the most recent tape even after dying, fast forward through the bulk of the level, and pause a few seconds before the end. Then they can take over playing again from there, knowing the perils of the misbehaving elements, and try again until they're more cooperative.