Added 3 files as two simple demo. Observe the results from the console. longtask-in-parentiframe.html doesn't log the longtask while demo-longtask-in-same-iframe.html do.

Deleted: demo-subframe-observing-longtask.html 454 bytes

demo-subframe-observing-longtask.html 454 bytes View Download

Deleted: demo-longtask-in-parentiframe.html 343 bytes

demo-longtask-in-parentiframe.html 343 bytes View Download

Deleted: demo-longtask-in-same-iframe.html 487 bytes

demo-longtask-in-same-iframe.html 487 bytes View Download

I'm unclear on what exactly the symptom is: 
I'd thought the issue is that a long task generated in the parent iframe is not observable by the child iframe (same origin), such as the case in this test:
https://cs.chromium.org/chromium/src/third_party/WebKit/LayoutTests/external/wpt/longtask-timing/longtask-in-parentiframe.html

However I made a simple repro and it works just fine:
https://panicker.users.x20web.corp.google.com/www/longtask/demo/repro747539/child-observes-parent.html

Max, in your demo the long task in parent is likely firing too early, before the observer in the child is ready (that's why I made it on-demand in my demo with a button)

Max, feel free to close the bug, unless you can demonstrate the issue

Shubhie, thanks for helping. I have tested using your sample and added a long task at the beginning, and you are right that the original test fires the long task earlier than the iframe is created. 

I think it makes sense that the long task will be filtered if there is no frame attaching to it. 

I should revise the test so that the long task being fired after the creation of iframe.