Finish the Incidental Count Sweep, Including Counted Section Titles #3169

Context

The "Incidental Counts" rule in GOVERNANCE.md "Documentation Style Conventions" (#3160, PR #3168) came with a sweep of the carried text. That sweep was a first pass. It left two groups of counts unchanged.

Gap

• Unjudged hits. The sweep grepped number words, "both", and "pair" across AGENTS.md, GOVERNANCE.md, CODESTYLE.md, WORKFLOW.md, .github/copilot-instructions.md, and .agents/skills/. Many "both" and "two" hits were never read in context. Where a wrapped skill line held a second count, only the first was checked.
• Counts in section titles. Some counts sit in a heading that other text quotes by name, so removing the count means renaming the heading and every reference to it. Constructed shape: a heading "Every case ends in one of four outcomes" that sibling files cite verbatim. The heading's own text then has to change in every citing file at once.

Possible Direction

Re-run the sweep with each hit read in context. Judge it by the rule: keep a count the reader needs, a pair fixed by construction, or a count an ordinal structure later in the paragraph relies on. Rename the counted headings together with every reference, using an untruncated slug search. A removal often leaves a later "both", "either", "the third", or a pronoun bound to the wrong noun, so read each edited paragraph whole afterwards.