This is yet another comparison between shmup scoring & movement games in order to justify my previous comparison that didn't seem to land. Maybe one day everyone will understand!
I love "double edged" challenges in games. The basic idea is that the very things that the players are threatened by, be it attacks, bullets, enemies, pits, should also serve a second purpose. Skilled players should somehow use those threats to their own advantage.
I think this creates a very rewarding learning curve and very dynamic, interactive gameplay because it allows the player's risk vs reward calculations to constantly and unpredictably change based on the situation.
However, designing games like this can easily lead to you, the designer, into being trapped in a bell curve of difficulty. There is always the danger that the "double edged" nature of your challenges will lead to a nasty feedback loop which outright prevents your game from becoming difficult past a certain point.
Psyvariar Revision is perhaps the best example of this trap in action.
To sum its mechanics - getting close to bullets without touching their hitbox earns you buzz, once you get enough buzz, you'll level up. That level up state gives you a second of invincibility, which can then let you "rub" even more bullets for even more level ups. The core gameplay of the game ends up being all about motorboating huge clusters of bullets, completely inverting bullet hell risk vs reward dynamics.
Despite the iframe chaining being dominant, the game derives its challenge from you getting hit by bullets. As a result, the hardest parts of Psyvariar aren't massive, dense bullet clusters but rather more mundane, simpler patterns. The game cannot make particularly difficult encounters or boss fights with its mechanics & the types of bullets/patterns it uses.
While Psyvariar is a really good, clear example of this problem in action, I think this problem pops up everywhere. For example :
BallisticNG (amazing game btw, play it)
Here's a game where this exact problem manifests itself in much more subtle, interesting ways. It's a clone of the PS1 era Wipeout games that (mostly) faithfully recreates their mechanics. One of its key features is being able to pitch your ship up & down to match the rather bumpy terrain of the tracks.
When starting off, the bumpy terrain can be pretty tricky to manage - if you don't pitch your ship correctly you'll slam into the floor constantly, slowing you down. Once you get the hang of the physics & build up speed (& learn its weird slingshot like airbrake physics), any bump becomes a ramp you can use to fly off-track. This turns what was meant to be a challenging element of the gameplay into something completely broken - you can now skip turns. Tight U-turns can be skipped, Chicanes might as well not exist, even entire sections of the track could be skipped.
This sort of skip means that BNG cannot have particularly difficult tracks which truly test your pitching. Because the very things that make them difficult (high speed, tight turns with varied elevation) also let players completely skip the challenges they're faced with. The designers are trapped in a "middle space" of difficulty, they can wiggle around but can't truly break out of it.
I think it's important to be aware of these sorts of feedback loops when designing a game's mechanics. Because while it might feel awesome to use threats as resources at first, it doesn't always scale well. It might force you to resort to more artificial, system-level fixes down the line.
Psyvariar's developers reduced EXP gain during boss fights and X levels to compensate for its iframe chaining, and BallisticNG's developers implemented a pretty complicated anti-skip system, and will now forever have to carefully make sure that any "difficult" part doesn't undermine itself and others like it when designing tracks.
If you wanna avoid this trap, think proactively! Make sure that no matter how much advantage the player builds, the challenge will always outpace them. Or separate the advantage from survival difficulty via the use of score. There are a lot of ways to dodge this trap, but it has to be considered early on.