Our jabber server takes its list of contacts and the globally shared roster from the LDAP
server, which requires nightly restarts of the jabber server for the lists to synchronise
with the LDAP. However, all Hudson instances we have fail to reconnect to the jabber server
after such a restart, unlike "regular" Jabber chat clients, causing (among other things)
build notifications to usually fail. It throws exceptions then.

Restarting the Hudson instances nightly (via cronjob) is out of question, in case they
were building or something, as a hack.

Please let the Jabber plugin retry connecting to the Jabber server every, say, minute or
so when the connection was closed.