What's been going on with CI recently? Seeing some flakiness.

Known sources: • timeouts of Python tools is likely from the changes to stabilize tool lockfiles, which added new tests like testing that every version of Python works w/ the tool. If it times out, please bump the timeout • JVM tests were restored recently but are still flaky. @fast-nail-55400 has been working on stabilizing. (Should we skip in meantime?) • macOS jobs sometimes time out, no idea why

I’ve been looking into the JVM test flakiness. Theory: The engine is running multiple copies of coursier concurrently, and unpacking the JDK download conflicts with other attempts to unpack the download by other coursier processes.

@fast-nail-55400 should we skip those in the meantime?

are they consistently breaking CI?

maybe 5% of runs?

I'm seeing

Failed to load Toolchain token from env var 'TOOLCHAIN_AUTH_TOKEN'

, that's just a warning, but still (cc @polite-garden-50641 when he gets back from vacation)

That's almost certainly not a flake

the token error is just a UX issue. I am following up on that https://github.com/toolchainlabs/toolchain/issues/10846

OK, that error is in fact due to my change interacting badly with something