OK - found what seems to be the problem. It seems a long-running query was locking up a db_handle (and a connection thread) every time it was run. Since I have more threads than handles, in the end db_handles run out.
Is there any way to monitor the number of handles in use (from the aolserver/driver point of view)?
And: Is it normal for queries to keep running in the postgres server after I restart AOLServer?
