On 7 June 2013 02:32, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Hm, good point. That reinforces my feeling that the page-number-based
> approach isn't workable as a guarantee; though we might want to keep
> that layout rule as a heuristic that would help reduce contention.
Can the locks just be taken in, say, numeric order of the pages involved?