If the database server disconnects there were exceptions
that were not trapped which would cause the recorder event
loop to collapse. As we never want the loop to end
we trap exceptions broadly.
Fix a bug in the new commit interval setting which caused
it to always commit after 1s