The duration of first render server side increased from 35ms to 70ms.
The DB is on the same node, just changed from 1GB shared cpu to 4GB and dedicated cpu.
So there shouldn’t be any network involved here.
I swear, the purple line used to twice as fast. And the server logs agree with me.
I’m having trouble believing it could be gettext adding 30ms and I don’t recall adding any new features in the past 3 months that could do this. Any hints on getting a profiler on the full scope of the initial page render? Otherwise, I have to do this the really, really old-fashioned way and start ripping parts of the page out / rollback / etc.
