Bugzilla

Here is the short list of things that I think we should consider doing:
1. Ensure that viaduct sends a lowercase content-length.
2. Assert that content-length is present and lowercase, probably only in debug builds, because that might be a little too expensive to ship.
3. Run a try build to see if those assertions flush out any other cases where this is being set directly.  More investigation might be required there.
4. Along with [this assertion](https://searchfox.org/mozilla-central/rev/435a77f1a1aaf1a78d30a2aaa81c6158a2f83dba/netwerk/protocol/http/Http3Stream.cpp#273), set `rv` to something other than NS_OK.  Having this function return OK when it was not is how we got into the infinite loop.

Here is the short list of things that I think we should consider doing:
1. Take emilio's patch.
2. Assert that content-length is present and in the expected casing, probably only in debug builds, because that might be a little too expensive to ship.
3. Run a try build to see if those assertions flush out any other cases where this is being set directly.  More investigation might be required there.
4. Along with [this assertion](https://searchfox.org/mozilla-central/rev/435a77f1a1aaf1a78d30a2aaa81c6158a2f83dba/netwerk/protocol/http/Http3Stream.cpp#273), set `rv` to something other than NS_OK.  Having this function return OK when it was not is how we got into the infinite loop.

(Updated to correct some mistakes)