It looks like this is because tests are running on infrastructure that's being made obsolete? But can't say for certain yet. I'd like to work on fixing this, so we can be sure PRs are doing what they ought to.
I disagree... these seem like build errors I've seen before where a specific branch fails to spool up. This is something that unfortunately happens with travis from time to time - you always need to look at the build log rather than rely on the checkmark
Hmmm, I'll leave this to you, but it's intermittent (so not code), and when I've seen it (recent mongo errors not withstanding) it has been due to a timeout booting the VM. Re-running the test (e.g. by pushing a non-code commit) works
There's also the other vm architecture they talk about, but when I looked that looked like a lot of work to convert and wouldn't be guaranteed to solve the problem (and would likely introduce a bunch of new ones)