For when you get back, I was under the impression that this already works correctly. Are there known discrepancies here?

Yeah, so I'm pretty certain that this should hold. Let me run some tests on try to see what happens if we make BrowsingContext::GetParent === GetSameTypeParent()

So I think that I filed this on your feedback for something. It should be correct as is, but I'm running a try with asserts, if that is horribly orange I'll reconsider otherwise this is just a wontfix:

So I think that I filed this on your feedback for something. It should be correct as is, but I'm running a try with asserts, if that is horribly orange I'll reconsider otherwise this is just a wontfix:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=afe096b3018881776bfd5bf7c1c72b2fa2109ba7

So the try from Comment 4 was pretty orange, so I'm running a test that asserts the other way around:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=ead6d4a7b71e8a545bb6e7f69215ee69ad441748

because I think that at GetSameTypeParent might be the same as GetParent at the time of GetSameTypeParent being called, but not the other way around. This would then also be because of how docshells are destroyed and how that differs from how browsing contexts being detached.

So this seems equally bad. Will dig into it to see exactly what's happening in the latter.

So, checking that when we do get a docshell from GetSameTypeParen, then that condition holds for the browsing contexts as well: https://treeherder.mozilla.org/#/jobs?repo=try&revision=ec672943b4c789c81c22f3ff2637c9ea1576eb0f

Yep, the last of the runs seems fine. The former two is the old issue with tearing down browsing contexts isn't the same as tearing down docshells.