For my future self or anybody else googling around I thought I would write up on my findings quickly.
To use native modules the main-file must be loaded from a script-tag with a type=”module” attribute on it. But it turns out that script[type=module] makes a different kind of request to the server (specifically the mode+credentials parts).
This means that when we try to match that request against the cache, with cache.match(event.request), it won’t match the cache-entry for the file added using the normal request-type or the string-based approach cache.addAll([‘app.js’, ‘module1.js’, …]).
The solution should be to manually match the request-type when adding the the cache. Big shoutout to @jeffposnick for pointing me towards this 🤘
But it turns out this doesn’t work perfectly either. So I actually ended manually creating the request-object in the fetch-listener to match the cache… this works, but doesn’t seem very pretty, and hopefully will go away as this matures!
Obviously this is pseudo-code, and you should come up with a better approach for when to actually create the request objects yourself, and you probably want to fallback to the network as well… (and other things)