From rafaelw, in https://bugs.webkit.org/show_bug.cgi?id=78639:
"It's pretty strange to call disconnect() and still *possibly* get one invocation of your callback. We should just clear any pending deliveries.
It's possible that script way want to know immediately prior to calling disconnect() *what* mutation records are pending delivery. If this happens in practice, we can add a method to synchronously retrieve & clear pending mutation records."
This is easy for WebKit to change, but want to make sure it makes sense and get it fixed in the spec first.
I agree we should cancel pending deliveries. addEventListener does the same thing FWIW.
I don't think we should bother with getting a list of pending deliveries unless use cases come up.
(In reply to comment #2)
> I agree we should cancel pending deliveries. addEventListener does the same
> thing FWIW.
You mean removeEventListener.
I don't have strong opinion. Either way is ok to me.
No need to cc me if I'm the assignee. On the other hand, this once it made me fix the bug a little quicker: https://bitbucket.org/ms2ger/dom-core/changeset/d5065a7d4427
Assuming the above means to empty the record queue per new terminology.