Steps to reproduce:
Technically this isn't a crash since the app won't start because of the single instance check, but for the user it is equivalent to "crash on start".

If gedit gets zombified by failing I/O (e.g. when working with documents on a sshfs share that fails for some reason), there is a deadlock situation:

- gedit cannot be killed because it waits for I/O
- share cannot be unmounted because it is in use by gedit
- *** another instance of gedit cannot be started because the single instance check kicks in ***

The single instance check should try (harder) to check if the existing gedit process is healthy, and treat it as non-existent if it doesn't respond. The timeout for getting the response shouldn't be too long, e.g. equal to the running WM's non-responding application kill timeout.

Stack trace:


Other information: