A different theory, which is hard to prove without debug symbols
to provide more detail in the trace, is that it might be looping
due to following a closed loop of btree page right-links. Which
shouldn't happen normally, but perhaps this index is corrupt.
Can you try reindexing the indexes of the COPY's target table?
Thanks Tom, since the reindex we’ve been unable to reproduce the problem again.
The big question is do you have any random thoughts on how a index may of gotten corrupted? The db lives on a two SSC ZFS mirror, so I’d not expect disk corruption.
Best,
Robert