Bugzilla

* `cookies/name/name.html` seems to just be using http cookies, so I don't even we need to worry about `SetCookieStringFromDocument` in any process. 
* I'm also not readily seeing any way for a set-cookie and broadcast from the parent to trigger parsing on the child.

So I'm guessing the parsing is occurring again because the we are taking the alternative pathway to the child's `SetCookieStringFromHttp`. 

IIUC, this pathway should only be used on a proxy-connect success, however Leander recently discovered that it's possible this function is called in child and parent (from at least a test environment). I looked at this a bit and found 1 way to get make the function run in both processes: use an ipc_unit test that uses `run_test_in_child()`. Maybe this test also elicits the race?

IMO this _looks_ like a failure to restrict ourselves to one code-path for parsing in a test scenario, but maybe it happens in the wild and maybe there is code that depends on it, like say if it was needed for other edge cases like non-fission builds (?). @Leander do you have any more info on how you were able to run the `SetCookieStringFromHttp`  in both content and parent with a single http cookie?

* `cookies/name/name.html` seems to just be using http cookies, so I don't think we need to worry about `SetCookieStringFromDocument` in any process. 
* I'm also not readily seeing any way for a set-cookie and broadcast from the parent to trigger parsing on the child.

So I'm guessing the parsing is occurring again because the we are taking the alternative pathway to the child's `SetCookieStringFromHttp`. 

IIUC, this pathway should only be used on a proxy-connect success, however Leander recently discovered that it's possible this function is called in child and parent (from at least a test environment). I looked at this a bit and found 1 way to get make the function run in both processes: use an ipc_unit test that uses `run_test_in_child()`. Maybe this test also elicits the race?

IMO this _looks_ like a failure to restrict ourselves to one code-path for parsing in a test scenario, but maybe it happens in the wild and maybe there is code that depends on it, like say if it was needed for other edge cases like non-fission builds (?). @Leander do you have any more info on how you were able to run the `SetCookieStringFromHttp`  in both content and parent with a single http cookie?