Exactly - I think we may have opposite understandings of what "zalgo" means in this case. We agree on the facts that the same callback may be called from either a JS or a non-JS context, but we differ on our understandings of what it means to not release zalgo given that fact. It sounds like you're saying that the callback's execution context should be stable/predictable independent of the circumstances that triggered it, while I'm saying that the relationship between the callback's execution context and the circumstances that triggered it should be stable. I don't think the former is actually viable, because the only thing that's stable is that there is a snapshot, but the actual snapshot is still unpredictable.

One point of clarification - my proposal would allow escaping your calling environment and getting access to the HostGetTopLevelAsyncContextSnapshot if you can listen to an event that you know will get called from non-JS. Is that a problem? I still see it as more consistent, so if this does pose a problem, them I'm at a bit of a loss as to the correct way to deal with it.