It's foremost a naming problem, FixedBufferAllocator doesn't hint that it is actually is a bit of a weird mix of a bump and stack allocator (IMHO if it would be a bump allocator it shouldn't have a free function at all, and for a stack allocator the free function should probably be called pop).
However both doesn't match Zig's expected alloc/free allocator interface, which is an interesting design challenge on its own.
Not at all. Viewed one way it isn't a problem at all (the user found and fixed the issue). Viewed another way, it is a flaw in the docs for FixedBufferAllocator that it offers a "free()" call but fails to make clear that this only works when freeing at the end of the allocated region.
However both doesn't match Zig's expected alloc/free allocator interface, which is an interesting design challenge on its own.