Richard Huxton <dev@archonet.com> writes:
> Things to check:
> 1. postgresql.conf settings match - different costs could cause this
> 2. statistics on the two columns (trn_patno,trn_old_date) - if they
> differ considerably between systems that would also explain it.
The different estimated row counts could only come from #2. I suspect
David has forgotten to run ANALYZE on the second system.
I agree that EXPLAIN VERBOSE output is not helpful...
regards, tom lane