When Syncing with Crowd fails due to a bad user it fails for everyone after that user, this should be more robust.

We came across this during a support case, but here's how we think it happened:

• JIRA 4.3+ set to sync with Crowd
• Crowd initially set the output lower case names and group names
• Memberships for users "Rob" and "Joe" get synced across as "rob" and "joe"

• Later we switch off the lower case output. Resync.
• Again, memberships for users "Rob" and "Joe" are being synced across, and JIRA does a lookup of the name "Rob", finds none and tries to add a membership entry for "rob" in "group" but can't add it due to the unique column key constraint of the membership table.
• The entire sync now fails, and does not continue.

• The original situation is an unlikely one to get into, but possible nonetheless, and it should be finding that match.
• The part that is causing us the most problems however, is that after it fails to insert this one entry, the entire sync fails. It is possible to miss out on 99% of the memberships being synced because of one failure, it should continue, perhaps letting us know what happened in the UI or the logs or both, but otherwise finish the sync.