[01:44:12] <Krinkle>	 turns out that when you have only a single data center, popular stuff is more popular, and local latencies are on average more local. nothing surprising, but it led me to add a few annotations in Grafana so that random high-level graphs have an explanation.
[01:44:21] <Krinkle>	 For example: Memcached latency for the "sitenotice" cache key group. 
[01:44:22] <Krinkle>	 https://grafana-rw.wikimedia.org/d/lqE4lcGWz/wanobjectcache-key-group?orgId=1&from=2025-09-01T00:00:00.000Z&to=2025-10-17T23:59:59.000Z&timezone=utc&var-keygroup=anonnotice
[01:45:34] <Krinkle>	 tagged with {operations} {performance} {mediawiki} which should cover 99% of dashboards that enable annotations by a tag
[01:46:19] <Krinkle>	 might be worth making part of the switchover runbook perhaps? At least the first day (services switch) and last day (eqiad repool)