In our workflows, our playground tests are failing a lot. I believe it might be because they really are running right on top of each other. They don't install a module, so they're not creating a server reload. Might they be creating the same Project somehow? Could they be starting at the exact same millisecond? The most recent tests for #582 did pass when I ran each test separately, even though they initially all failed. 2 of the 3 jobs (don't ask why there were 3) failed because of timeouts trying to get to interview pages. The other one failed because it couldn't find the translation spreadsheets. I'm not sure why that would happen either.
We can check if they made the same Project by checking the project name in each, which we do log with "Project called"
I've added a 60 second delay to the PR workflow and the tests still fail. We're going to just dedupe the workflow triggers. If any future person discovers why this would be happening, let us know!
In our workflows, our playground tests are failing a lot. I believe it might be because they really are running right on top of each other. They don't install a module, so they're not creating a server reload. Might they be creating the same Project somehow? Could they be starting at the exact same millisecond? The most recent tests for #582 did pass when I ran each test separately, even though they initially all failed. 2 of the 3 jobs (don't ask why there were 3) failed because of timeouts trying to get to interview pages. The other one failed because it couldn't find the translation spreadsheets. I'm not sure why that would happen either.
We can check if they made the same Project by checking the project name in each, which we do log with "Project called"
Tests 1, 2, and 3.