That's because they're just <h2> as opposed to <h2><span id="mw-headline"> - we're converting only the span, never touching the h2's themselves. What do you propose to do with them Jon?

Any reason these don't contain spans?
Sounds like something to change in core to be consistent with other headings.

Because they're not result of wikitext parse.

Any reason we can't split on h2 with optional span?

Hmm, I can *theorize* that this behaviour is due to us originally screen-scraping everything, including headings from chrome.

One thing we could do is transform all the HTML on the client and drop headingTransform altogether... this scares me but in theory it should be possible...

We /could/ do this in JavaScript. This would involve collecting known headings and constructing the sections manually to collapse them.

https://gerrit.wikimedia.org/r/#/c/83837/ fixes this.

This should now be resolved. yey!