akaster: in general, it seems like null client fetches are not stress-tested. (annevk has expressed a similar opinion.) I filled in fields based on https://fetch.spec.whatwg.org/#fetch-elsewhere-request (originally written by Noam Rosenthal) which says you need "leave request’s client as null and set the request’s origin, policy container, service-workers mode, and referrer to appropriate values instead". It's very possible that list is not exhaustive.

However, I'm unsure whether we should fix it in fetch or in all call sites (e.g. in HTML).

Hello, I have a clarification question concerning the HTML Tokenizer;
The "named character reference state" states that characters consumed as part of a character reference should be flushed if there was no match. However, it also requires to "Consume the maximum number of characters possible, where the consumed characters are one of the identifiers in the first column of the named character references table" beforehand.

I'm a bit confused on what this is supposed to mean. Should it only consume characters (as in pop off the incoming stream) if they have a matching entry in the spec table or unconditionally? In the latter case, how much characters would need to be consumed?