Right now we just spinelessly return the default computed values in this case. Turning this into an assertion would probably reveal a number of bugs, which might otherwise be harder to track down.
Over to heycam. If this trips a lot, prioritize diagnosing the issues and filing bugs. First priority is finding out how much work is hidden under this one.
Looks like there are around 40 tests that trigger that assertion in Servo_ResolveStyle: https://treeherder.mozilla.org/#/jobs?repo=try&revision=204196e5f095a5eefa24188e130334b67feed300&group_state=expanded If I make Servo_ResolvePseudoStyle also assert, I get a bunch more, but I'm not sure whether it's right to assert there. (Is it possible we're probing for pseudo styles before we resolve styles for the element?) https://treeherder.mozilla.org/#/jobs?repo=try&revision=afefbf844d947d8a4ba161467e90edc632b41437&group_state=expanded
Let's scope this issue to non-pseudos and file a followup for the pseudo stuff, since it's still a bit busted until I get time to land bug 1331047.