On Tue, Apr 16, 2013 at 6:09 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
[...]
Or, if you'd rather a more direct answer: wanting this sounds like evidence of bad application design. Why is your app dependent on getting failures from currval, and isn't there a better way to do it?
The sequence cache (seqtab) is used per each backend, so if we use a connection pooler (like pgbouncer in session mode) between our app and postgres we can get a wrong value from 'currval' because the backend isn't completely clean.
This isn't it a good reason to implement this feature?