I do have a case where in dev mode the app loads heaps of single resources (css/js/etc). Most of those are rather small and in prod they are combined, compressed, etc.

Since the profiler uses uniqid() to create the profile token I see randomly (but consistently) warnings about files/dirs that already exist (using FileProfileStorage). I would expect other storage classes to suffer similarly.
This seems to be a racing condition where two requests do have the same request time in milliseconds.

Are there any reason to not use the more_entropy flag and/or the prefix on the uniqid() call?

Alternatively, are there other ways to avoid this by different means?