Martin Ginkel writes:

 > But this is only the theory. In practice for me this fails in at
 > least 30% of the cases. especially for server-server conflicts, where
 > you have no alternative to repair.

I've never used more than one server, so I can't speak to
server-server conflicts.  I'm not surprised at your experience; as I
said I find repair to be pretty buggy.  My point is simply that the
functionality provided by repair is *necessary* if you want a means of
resolving conflicts short of reinit.