We have a time-based partitioning scheme and with Postgres 8.4 on RHEL 6 we would "detach" a partition that was no longer being written to and CLUSTER it before "re-attaching" it [1]. This never seemed to have any issues, but we recently upgraded to Postgres 9.2 on RHEL 7 and we've noticed a few instances where prepared statements fail and it appears to be caused by running NO INHERIT to "detach" the partition that is no longer being written to. Is this an issue caused by the changes in the planner in regards to how prepared statements are handled [2]?