~~When we move the mobile performance device lab to our new provider, we should make sure developers easily can run their own tests on demand.~~
We should make sure it's easy to run performance tests against different URLs/user journeys and compare the result. Today you need to commit the test to Git and then let our server pickup the tests and run the tests over and over. If we could make it possible to run on demand, we will make it a faster for us to know if a change has a potential performance regression or not.
We could potentially do so that devs can trigger a test (https://github.com/wikimedia/performance-synthetic-monitoring-tests) from Slack/IRC/web and then get a link to the result when the runs are finished.
I've started to implement this some time ago (modernising the work with the scholarship I got from the Swedish Internet Foundation many years ago) and I have a working version but what I wanna do before we can use it making it possible to change backend (today agent/server talks to each others using Redis, and it should be possible to implement a interface so we can easier can change backend).