When using Page.navigate we don't return a proper loaderId when the target url does not start with http/https: https://searchfox.org/mozilla-central/rev/7f729f601c0b738f870ae0ed49098f9268e250f9/remote/cdp/domains/parent/Page.jsm#107-110

Because of this, the current implementation of page.goto in puppeteer actually dismisses the event. Navigation works by accident because we incorrectly emit navigatedWithinDocument which tricks puppeteer into thinking that the navigation was a "same document navigation".

However we want to stop emitting those incorrect navigatedWithinDocument events and therefore we should fix navigate to always return a proper loaderId.

Note that this will be irrelevant for puppeteer once https://github.com/puppeteer/puppeteer/pull/8369 (re)lands, because loaderId will no longer be checked, but we need to ensure minimal backward compatibility with the current puppeteer so we can't just ignore the loaderId issue.

This is basically a dupe of bug 1632008. All about:* pages are actually affected.