This actually doesn't work, and may require changes on the HTML side. The problem is that we can return from the above algorithm without starting a new navigation, and this step will get the id of some other navigation (or maybe null). Previously that was OK because we already knew the navigation id at this point, so in the next step we'd notice that a failure event had been dispatched and return. But now we need the navigation to be initiated to find the correct id.

I wonder if the part using await for events after the navigate is also flawed, since by the time the await is called some events might have already happened?

Hmm, looks like it probably is because await doesn't have a way to synchronously check if the resume condition was already met.

I am not sure I am ready to propose a change to the HTML spec right now, but perhaps we could specify here that we want to yield the navigation id back to our algorithm as soon as it is defined on the step 6 and run the rest of the navigate/reload algorithm in parallel?