Risk register is ID-ordered while chapter 11 promises priority ordering

[dieterbaier]

Problem

src/docs/arc42/doc-11000-risks-and-technical-debt.adoc tells the reader:

Risks are ordered by derived qualitative priority.

The generated register is ordered by artifact ID.

Evidence

src/docs/arc42/11-risks-and-technical-debt/generated/doc-11001-risks.adoc, in
render order, with the Priority column each row carries:

Row	ID	Priority
1	R-001	High
2	R-002	High
3	R-003	Medium
4	R-004	Medium
5	R-005	Low
6	R-006	Medium
7	R-007	Medium

A Low priority sits above two Medium ones. The sequence is exactly ID order,
which coincided with priority order until R-005 was added.

Pre-existing on main. Appending R-007 in ID order neither causes nor worsens
it, but it does add a second row below the misplaced one.

Decision to make

Two honest resolutions, and they are not equivalent:

• Sort the generator output by priority, with a stable tie-break on ID.
  Keeps the promise the chapter makes, and puts the risks a reader most needs
  first. Changes the order of an existing rendered table.
• Correct the chapter prose to say the register is ordered by ID. Cheaper,
  and honest, but gives up a property worth having in a risk register.

The first is the better outcome if the sort is cheap; general-semantic-contracts.md
asks chapter 11 for priority ordering, so changing the prose narrows what the
contract promised.

Behaviour specification

Sorting changes observable generator output, so bdd-specification applies:
a scenario in features/documentation-generation.feature asserting the register
order for a set of risks with mixed priorities, bridged into the generator
tests. Prose-only correction needs no scenario.

Acceptance criteria

• The chapter prose and the generated register agree.
• If sorting is chosen: ties break deterministically, so the output stays
  reproducible; a scenario covers the ordering and is bridged to a test.
• If prose is chosen: the contract's expectation for chapter 11 is revisited
  in the same change rather than left contradicting the prose.
• Render and validation checks pass.

Origin

Found on 2026-09-04 by an independent Convergence Check on #88, which reported
it as a pre-existing residual observation rather than a defect introduced by
that change.