Deployment tests are still flaky

[TimMoore]

These seem to fail almost always on the first run, but usually pass on a re-run. I suspect a race condition.

We might be able to mitigate this by limiting to one concurrent job, at the expense of slowing down the happy path. WDYT?

[ignasi35]

We might be able to mitigate this by limiting to one concurrent job, at the expense of slowing down the happy path. WDYT?

Sounds good to me. I haven't looked into the deployment tsts in a while so really hard to guess. Reducing the concurrent number of jobs is very cheap so let's test that, yeah.

[TimMoore]

OK done. I also added daily cron builds for the 1.5.x and 1.6.x branches.