Great feedback!
I had some ideas similar to what you are suggesting, but the time constraints of the game jam prevented me from adding too many unique mechanics. GB Studio is a very clunky, clearly work-in-progress piece of software that is honestly a bit of a puzzle to work with, and I was learning how to use it as I worked on this.
My enemy movement code is pretty inefficient, and enabling collision on their pathfinding makes it unplayably laggy (even in GBC mode). Allowing them to walk through walls was a necessary sacrifice because I didn't have time to fix this. With more time I probably could have optimized my enemy scripts and made it work with collision.
The sprite garbling most likely happens because I am close to the sprite limit in the scene. Technically there are 6 enemy actors, and each one redundantly loads a copy of the spritesheet for BOTH enemy types - this is actually gbstudio's fault, I'm not quite sure how to do things dynamically and efficiently within gbstudio's clunk. Might require some mixture of GBVM and ASM.
I'm still debating whether or not I would like to keep working on this and polish it into a complete project, but I had tons of fun working on it & the feedback I have gotten does seem to imply people see potential in it.