Could one transaction (one that should be relatively simple and short) cause another complex, long running transaction (involving INSERTS, on a table the first transaction may be reading from) to take many orders of magnitude longer than it would normally? (short of competing for system resources, like CPU time etc, of course)
I don't believe my scenario involved a deadlock but I expect my short transaction was probably blocked by my long one. Does it make any sense that this could very significantly affect the performance of the non-blocked transaction?
Thanks,
Eric
Have you tried wait event analysis (looking at wait_event_type, wait_event, state, query samples from pg_stat_activity)?
