> I think that we could get rid of the initial_restart_lsn and just use > s->data.restart_lsn here (while keeping initial xmin ones to preserve the > intent of 818fefd8fd4 for those).
I agree with your proposed change.
> Your concern is only about the restart_lsn, right? (asking because I don't think > it's safe not to rely on the initial_* for the xmin ones, see [1]).
My primary concern is indeed the restart_lsn issue, and thanks again for the clear explanation.