When calling Page.setLifecycleEventsEnabled, it looks like all events get flushed, including one called commit. This correspond pretty well to nsIWebProgressListener STATE_STOP.

Puppeteer depends on commit here: https://github.com/GoogleChrome/puppeteer/blob/d17708ba1fc8411941bfd07f633aaf2fb5cb671e/lib/LifecycleWatcher.js#L156

See also https://bugzilla.mozilla.org/show_bug.cgi?id=1597253#c4

I don't think this blocks gutenberg tests at the moment.

Maybe we have to indeed switch from handling the individual navigation events (DOMContentLoaded, pageshow) to web progress listeners.

It may also mean we have to always listen for the events internally
but avoid emitting them until the client enables the domain, so
that we can build up a cache to flush.

Chrome actually retrieves the timestamps for each of those events on the current loader. As such it is able to use those from the last navigation. I wonder if we have a reference to the loader, and could do the same.