I read the Structured Headers for HTTP proposal and browsed the test suite. Looks reasonable.
As someone who has written a few custom parsers for HTTP headers, I welcome this.
Agreed with at least "worth prototyping".
With the potential to reduce duplication of code, and especially reduce infrequently-run code paths for rare(r) headers, I think there are likely security benefits as this is adopted for new (and possibly old headers).
For that reason I think Mozilla should consider this as "important" or at least state that we can see this being/becoming "important".