Comment 88 for bug 332270

Having left the problem alone overnight it occurred to me this morning that maybe the issue here is the *multiple* watches being created by similar devices (block + lvm, block +md) in the same event run that is at the heart of this. This would explain why only systems with some combination of LVM, crypt, md or dmraid seem to be affected.

With my earlier observations about timing being significant (slowed down by udev logging, it can avoid the issue) I wonder if it's a case of the actions in 65-persistent-storage not having completed when 85-lvm2 (for example) is being processed and therefore creating a positive feedback loop?

From what I can gather child processes are forked so we could have a child processes triggering new inotify events unexpectedly.

I did wonder briefly what the effect is when an inotify watch is created on a node that is already covered by another inotify watch.