As many of you have noticed (and pointed out to us!); Confluence is very temperamental.

We’ve narrowed this down to “something” in Resin (our J2EE server).

We’ve also prodded Contegix into changing how Confluence is monitored so it can detect stalls (which it previously ignored some of the time).

Hopefully this will mean that there are less stalls (you might notice more breakages though as the server is bounced).

We are still working on making Confluence reliable however.