So I think I’ve run into this before when trying to help folks get bind mounts to work. Take a look at this little adventure I went on (no sound):
It appears that a simple container lifecycle event is enough to induce “trouble” in that volume. (I’m currently using gRPC FUSE, btw). Solution was as you describe–restart Docker Desktop (and thusly the underlying VM+Docker Engine).
EDIT: So this does feel like a bug in Docker Desktop. I’m sure you could replicate this same behavior in even a basic base image with a similarly configured mount.