The following process always leaves a stale lock file:

1. darcs record in repo A, leave it in the state of editing the patch  
description
2. darcs push from another computer to repo A, leave it in the staet  
of "waiting for lock"
3. finish the patch description in repo A

Now the push proceeds, but in this case I get "Apply failed!" (is  
that the way that darcs-2 reports conflicts?).

Now there is a leftover lock file in repo A.

So actually maybe there is a bug here and darcs-2 is not intended to  
write "Apply failed!" and leave a lock file behind...

Regards,

Zooko