The idea is interesting and in line with the jam's theme, but units switching allegiance based purely on randomness makes performance mostly up to RNG despite of being a tactics/strategy style game. Perhaps had there been more time, better criteria could be set for units switching sides. RNG can be a factor, certainly, but it's a bit much if it is the only factor. Also, tactical/strategic implications are mostly moot when the units are all set in a confined cluster, as player choice just boils down to doing what's available from RNG each turn.
Key-based controls feel a bit unintuitive compared to mouse or controller here, and it took a while to figure out exactly the sequence of confirmation inputs for some of the actions. For the UX/UI, a tile-range highlighting display a la Fire Emblem games to show movement and attack range would have been nice as well, especially with archers not being able to shoot over allies to land ranged hits against distant enemies.
The core gist of the idea is interesting and could lead to some inspired mechanics and emergent gameplay, but the execution on it here falls short of that. That said, to capitalize on more promising possibilities with the idea would probably require more time and resources than the jam constraints are friendly to.
The art and sound effects are nice, and the BGM suits this type of game and setting. Unfortunately, I can't really give a higher presentation score, seeing as these are pre-existing assets.
In all, a valiant attempt at something new and different, but this iteration kind of misses the mark. Good job completing a game project in 48hrs, and good luck iterating if you decide to explore the core idea further.
P.S. The restart menu that comes up at the end of a game doesn't seem to respond to either mouse or keyboard inputs for me. Not sure if it's a bug, or perhaps I'm missing something.