I think I've fixed it. As I wrote in this previous post, feedback from the first playtest was that the movement cards were too restrictive and it made the game really frustrating. The first thing I did after that was come up with a really simple AI so I could solo test, playing as the Faeries against a dummy Villager player. I ran through one game using the original rules to get a baseline, and found that on about half of my turns I was able to make a move that felt strategic, as opposed to just filler or even not having a legal move at all. Then I tinkered with some different ideas and tested those out. The winning idea is to keep the original deck of movement cards as "special moves," with all of that working the same way as before, but also each player gets an identical set of "basic move" cards. The basic moves are set out face up in front of the player, and on your turn you can either use any two basic moves that you have available, or one special move from your hand. When you use basics you flip them face down, and they only get turned face up again after you've used a certain number of them.
In my one solo test (so far) of this system, every single move that I made as the Faeries felt strategic; even if it wasn't the best move I could imagine it still felt like it was leading somewhere. And apart from the statistics, just the play experience felt like I did a lot better than before with finding the right puzzle challenge level. Hopefully other testers will agree the next time I get to break this one out in public.
Next step - run enough test games to see if the two sides are balanced, and figure out how to fix it if they're not. This could take a little while...