@DStrine there is only one cause that is semi-unresolved. Maggie did a work-around that means it is not failing now but we still have occasional discrepancies on is_primary from deduping which I think I can replicate in a test & if we can reliably solve that we can removed Maggie's work around (which would be better from a data-integrity point of view but at the margins - ie there is a risk that an email could wind up on the suppression list that shouldnt but I think the volume is maybe a handful a year & we have other ways emails can leak onto the suppression list

@spatton and @MNoorWMF and @KHaggard FYI I think a few people accidentally added to the suppression per year would be an ok price to pay but looping you in here.

Thanks for looping, David! I think for data integrity's sake, what Eileen said makes sense. I'm just curious (no urgency) - can we quantify approximately how many or what percentage of the list this might happen to?

@MNoorWMF - my guess is maybe a few hundred a year between this & other gotchas. That's what I thought the purge & rebuild of the suppression list occassionally was supposed to clear out

I've moved this into done as they have pretty much stopped after the last patches

This covers the duplicate issue https://phabricator.wikimedia.org/T255165 also https://phabricator.wikimedia.org/T256096