UUIDs have great characteristics as well: they don't leak information about the size of your data (or the rate it changes,) they are easily masked, you can have

UUIDs have great characteristics as well: they don't leak information about the size of your data (or the rate it changes,) they are easily masked, you can have clients generate stable ids (which you hint at in distributed systems.) The db performance degradations you mention are highly implementation-dependent.