renderedText substitutes a space for a removed HTML tag so a tag inside a word splits it and the excluded heading escapes on all four heading paths

renderedText substitutes a space for a removed HTML tag so a tag inside a word splits it and the excluded heading escapes on all four heading paths

Found by the round-9 adversarial security review of build/itd-183. REGRESSION INTRODUCED BY 2225d6cb. The cheapest of the round's regressions to trigger, and it defeats all four heading paths at once.

renderedText now does out = htmlTagRe.ReplaceAllString(out, " "). A tag INSIDE a word therefore splits the word, and the slug stops matching:

renderedText("Audi<i>t</i> Notes")
  HEAD:   "Audi t  Notes"  -> slug audi-t-notes
  parent: "Audit Notes"    -> slug audit-notes

So ## Audit Note<sup>s</sup> in any admitted file passes the section loop's exact-title lookup, then passes namesExcludedHeading -> sameRendering, and the section travels. No HTML-block trickery and no unclosed anything is needed; a renderer shows exactly "Audit Notes".

Verified admitted on HEAD and refused on the parent on every path: ## Audi<i>t</i> Notes (section scan), ## Audi<i>t</i> Notes (indented ATX), Audi<i>t</i> Notes\n=== (setext), <h2>Audi<i>t</i> Notes</h2> (raw HTML), and the same shape against Why This Matters and Open Questions.

The round traded ONE case (Audit<br>Notes, which the parent leaked) for SEVEN. This is the exact pattern the eight prior rounds were trying to break: closed on one path, opened on its neighbour.

Remedy (class-closing): take BOTH readings, exactly as this same commit already did for the soft heading bound. Compare the slug of renderedText with tags replaced by "" AND with tags replaced by " ", and refuse if either names an excluded heading. That closes Audit<br>Notes and Audi<i>t</i> Notes together and cannot reopen either.