[00:11:58] <AaronSchulz>	 Krinkle: right
[00:12:26] <AaronSchulz>	 the /rpc job runner was an HTTP POST that ran many jobs (not just one)...but even then, JobRunner explicitly runs deferred update
[00:15:28] <Krinkle>	 AaronSchulz: right, so JobRunner already handles it proactively.
[00:15:41] <Krinkle>	 AaronSchulz: probably not for stats buffering though?
[00:16:21] <Krinkle>	 not as big a deal but also unchanged either way, so I'll just leave that as-is
[00:22:02] <AaronSchulz>	 not a big deal...would be nice to integrate into DeferredUpdates (like I did with JobQueueGroup::lazyPush) and be recursion-proof (in case of stats for queued updates)
[00:24:58] <AaronSchulz>	 Krinkle: you can check if MEDIAWIKI_JOB_RUNNER is defined
[00:25:34] <Krinkle>	 AaronSchulz: I don't want to distract from sqlbagostuff for multi-dc but if you have a minute at some point today/tomrorow, I've confused myself thouroughly by https://gerrit.wikimedia.org/r/c/mediawiki/core/+/679984 - could use a fresh pair of eyes to get me back on track. 
[00:25:41] <AaronSchulz>	 the rpc/ runners do that, along with runJobs.php
[00:25:59] <Krinkle>	 ack, yeah, I might hack that into emit stats buffer as an improvement there
[00:26:30] <Krinkle>	 (later)
[18:03:25] <Krinkle>	 I've shuffled it around for now such that the Setup code for shared DB stays around until the end of the chain. That doesn't affect most setups (both local, CI, and prod) anyway.
[18:03:43] <Krinkle>	 allows the other more impactful ones to go first