Makes sense to me, NI'ing edmorley to see if there's some reason we can't/shouldn't do this.
We've talked in the past about fixing the remaining OrangeFactor deficiencies (I forget the thread or bug, but these include things like not supporting intermittents in builds [as opposed to tests], lack of stats and a few other issues; OF v2 using Treeherder's API will fix many of them) and then switching off the bot comments entirely, in lieu of periodic automated comments on the top N failures only. eg: "Hi! This bug was N out of Y this week and occurred 84% of the time on Android x86. It first occurred on date Z and the failure rate this week is 8 times last week. Here is a link to the full OF stats page: <link>" These would both give more useful info, avoid bug spam, and mean people are only pinged on the most important issues, so as to avoid bot-comment fatigue. However, until we have the above, people have tools and workflows that actively rely on both there being a 1:1 mapping of comments to occurrences, and also the comments themselves remaining parsable (bleugh). I'm happy to morph this bug into something about the above though? (Can't remember if we have one on file; will leave the priority unset for this bug, so it stays in the triage queue, which I'll take a look at once I've caught up from 16 days of PTO/public holiday email).
(In reply to Ed Morley [:edmorley] from comment #2) > (Can't > remember if we have one on file; will leave the priority unset for this bug, > so it stays in the triage queue, which I'll take a look at once I've caught > up from 16 days of PTO/public holiday email). I don't remember seeing an actual bug for that plan, just comments in other bugs (bug 1069560 comment 0, for instance) about the eventual plan.
Ah well remembered - I'd forgotten about that comment :-)
There was bug 856658 in the TBPL component.