These were previously split up because the test methods were slow individually that they massively extended pipeline times. However improvements to PerformanceTestBase mean that we're able to make multiple performance test assertions within a single test method now.
This should reduce the load on pipeline tests (i.e. instead of running three test methods that have to install umami, we'll install one), but more significantly, it should help to reduce the current performance pipeline down from just under 9 minutes especially because performance jobs run the tests without concurrency, so they're end to end.
If wen can get the performance test jobs down under 7.5 minutes each, that will allow us to fit 8 runs into an hour, which in turn will make https://gander.tag1.io/?orgId=1&refresh=30s&var-version=11.0.2-dev&var-t... work for two branches at a time (because grafana doesn't successfully graph results unless there are more than three test runs within an hour at the moment.
Locally it goes from:
Time: 02:45.601, Memory: 184.00 MB
OK (3 tests, 35 assertions)
Down to:
Time: 01:40.952, Memory: 184.00 MB
OK (1 test, 33 assertions)
Needs work
11.0 🔥
Well it nearly works but at least one CSS file isn't created by the time we try to check the size, which isn't good.