Hi,
Today, we had another replag problem, caused by Aka's user stats script. Replag kept rising even after the script was stopped (not killed). Apparently, the reason lies with the transaction mode used: a long running query may lock (parts of) the tables it uses, denying any updates the replication daemon wants to make.
the script already starts with
"set session transaction isolation level read uncommitted"
as told by River on this list a few months ago. Beside this, the last time it run it did not affect the replication lag. So, I'm not sure that this script caused these problems.
-- Aka