@pganssle I can break this up in several PRs; let me know what works best for you.

🤖 New build scheduled with the buildbot fleet by @erlend-aasland for commit 573aec1 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

You've beat me to it :)
Thanks!

Sorry, I was not aware that you were working on a PR; I should've asked first.

🤖 New build scheduled with the buildbot fleet by @erlend-aasland for commit 6e69e93 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

Thanks for the review, Kumar. I'll wait until Monday with landing this, to give @pganssle a chance to chime in.

When you're done making the requested changes, leave the comment: I have made the requested changes; please review again.

Hopefully the tests I put in place for the cache stuff are robust 😅 That kind of thing is hard to test well, and this is the kind of change that really needs it 😛

Yeah, I know, changes like this need really good test suites! I'll take a look at the coverage before and after the change, just to see that we've at least got all the branches covered.

Thanks for the review 🙂

Good to go, @pganssle?

I don't think it needs to block this particular PR, but is the subinterpreter work advanced enough that we can add an actual test that the module isolation worked yet?

I'm guessing the answer is no because we need to import datetime to get the timedeltas, and datetime is not isolated?

I don't think it needs to block this particular PR, but is the subinterpreter work advanced enough that we can add an actual test that the module isolation worked yet?

Subinterpreter testing is not optimal yet. There's some machinery in test_embed / Programs/_testembed.c. See also #98627. Thanks for bringing that up.

I'm guessing the answer is no because we need to import datetime to get the timedeltas, and datetime is not isolated?

Correct, _datetime is not isolated yet; I've got a fairly up to date proof-of-concept patch for it, though.

Thanks again for the reviews, Kumar and Paul; highly appreciated.