Hi, falken@, shimazu@, I'd like to take this, do you think this is a valid task to do now? Or some hints? Thanks.

Actually I'm confused about this behavior. 
My understanding is that the SW with a scope 'worker_script' should be responsible to intercept fetch events fired from the worker 'worker_script', but the test asserts that the SW should intercept the fetch for 'worker_script' script itself? Is there spec describing such behavior? Thanks!

Thanks for discovering and filing the bug.

Yes that's what the bug is about. It wants new Worker(in-scope-url) and new SharedWorker(in-scope-url) from an uncontrolled client to be intercepted by an in-scope service worker.

The spec seems to agree. A "non-subresource request" includes not just navigations, but also requests for worker and shared worker.
https://fetch.spec.whatwg.org/#non-subresource-request

Handle Fetch is supposed to handle non-subresource requests:
https://w3c.github.io/ServiceWorker/#handle-fetch

Chromium's current implementation just intercepts navigations and resource requests from controlled clients.

However, I don't think it's a good time to work on this. Our interception code will change quite a bit in  issue 715640 . So I'd hold off on big changes in this area for now.

But I do think that in general the bugs at issue 678905 are good things to work on.

> The spec seems to agree. A "non-subresource request" includes not just navigations, but also requests for worker and shared worker.

Agree, FWIW.

> Chromium's current implementation just intercepts navigations and resource requests from controlled clients.

Gecko has the same bug, but I'm working on fixing it.  Hopefully in the next few months.