Bugfix/database_rollback
Bugfix: Commit right away: timeShift() UPDATEs every message, and keeping those row locks until tic() returns blocks any other node that processes (and deletes) messages meanwhile. That node may hold timerservice_lock while it retries, so this loop could not process anything either: a deadlock until the request/test times out.
Was leading to sporadic occurrences of massive database rollbacks during live testing.