The various multi-data center improvements for T88445 are great but we need to make sure that people are writing new code the Right way too.
Need a good page on mediawiki.org that explains these issues in one place, plus update various other docs to make sure they're recommending the same things.
A key issue is making sure it's easy to queue up read-write background jobs from read-only web requests.
Another might be how devs can identify problematic code. Can we enforce read-onliness on GET through a configuration switch during development, throw warnings, etc?
Also, ensuring that hit storms to a single operation don't cause the same job to get run a hojillion times. (Deduping in job queue?)
There's probably a lot of information that can be pulled from both T88666: RFC: Master/slave datacenter strategy for MediaWiki and the IRC log for the RFC meeting where it was approved (E54).
Is mw:Requests_for_comment/Master_%26_slave_datacenter_strategy_for_MediaWiki a document we can use as a starting point for the "best practices" doc?