Item 31: close Scenario 25's demo-doc quote-count shortfall - #424

Closes out CLAUDE.md Active Backlog Item 31 entirely -- all 7 originally
flagged scenarios (2026-08-08 audit) now meet the 5-quote house-rule
minimum:

- "Reactive-only failure hint" (Part VIII): its own section title
  promised coverage of "both Tier A and requirement 9's real-build-
  failure paths" but only ever quoted requirement 9's. Added the
  distinct Tier A (:try_nuitka_tier_a) failure messages from source
  (both call sites, [Extrapolated Branch] since no CI run has hit a
  real Nuitka compiler failure) -- the section now actually covers
  what its title claims. 5 quotes total (was 2).

- Scenario 26 (Conda base periodic update): added the "last update <
  30 days ago" skip branch -- a real, deterministic literal string
  from source, not previously documented at all, and in practice more
  common than the "first install" skip already shown (it's what fires
  on every ordinary repeat bootstrap within the 30-day window) -- plus
  the update-failure line, previously only described in prose. 5
  quotes total (was 3).

- Scenario 41 (Interactive verification): added the PID-display line
  and the live-teed program's own real stdout (hello-from-stub),
  cross-cited from an earlier real capture in the same doc (run
  30328748330, job 90179708091) that already confirmed both as
  genuine console output belonging to this exact live-tee mechanism.
  5 quotes total (was 3).

Item 31 is now fully resolved -- moved from CLAUDE.md's Active Backlog
to docs/agent-closed-backlog.md's Closed Active Backlog Items section
(keeping its original number), consolidating the full three-PR
resolution history (#423/#424/#425) in one place.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW

…g bug (#425)

* Item 31: close Scenario 23/36's quote shortfall, fix real ordering bug

While sourcing additional real quotes for Scenario 23/36 (both about
HP_PVW_KNOWN_IDEMPOTENT execute-mode discovery), traced the actual
call order in run_setup.bat and found the doc's own ordering claim was
wrong: it said the discovery run fires "right after entry selection
returns," but :determine_entry is genuinely called TWICE -- an early,
silent pass right after provider selection (which
:pvw_known_idempotent_run's own gate depends on), and a later one,
after the entire dependency-install phase, that's the one which
actually echoes "Chosen entry: ...". Confirmed directly against the
real captured job log (run 30328748330, job 90179708109, uv lane):
discovery fires right after "[BOOT] REQ-009: Selected Python provider:
UV.", well before any entry announcement.

Both scenarios now quote the provider-selection line before and the
real "Chosen entry"/"Entry selected" pair after, both past the 5-quote
minimum and both now factually correct about the ordering.

Updates CLAUDE.md Item 31's tracking (3 shortfalls remain, was 5).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW

* CLAUDE.md: fix Scenario 36 quote count (5 -> 6)

CodeRabbit finding on PR #425: Scenario 36's console block has 6
lines (4 in the discovery block + 2 in the Chosen entry block), same
as Scenario 23, not 5 as the tracking note said. Confirmed by direct
recount against the current file.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW

* Item 31: close remaining quote-count shortfalls, move to Closed Backlog

Closes out CLAUDE.md Active Backlog Item 31 entirely -- all 7 originally
flagged scenarios (2026-08-08 audit) now meet the 5-quote house-rule
minimum:

- "Reactive-only failure hint" (Part VIII): its own section title
  promised coverage of "both Tier A and requirement 9's real-build-
  failure paths" but only ever quoted requirement 9's. Added the
  distinct Tier A (:try_nuitka_tier_a) failure messages from source
  (both call sites, [Extrapolated Branch] since no CI run has hit a
  real Nuitka compiler failure) -- the section now actually covers
  what its title claims. 5 quotes total (was 2).

- Scenario 26 (Conda base periodic update): added the "last update <
  30 days ago" skip branch -- a real, deterministic literal string
  from source, not previously documented at all, and in practice more
  common than the "first install" skip already shown (it's what fires
  on every ordinary repeat bootstrap within the 30-day window) -- plus
  the update-failure line, previously only described in prose. 5
  quotes total (was 3).

- Scenario 41 (Interactive verification): added the PID-display line
  and the live-teed program's own real stdout (hello-from-stub),
  cross-cited from an earlier real capture in the same doc (run
  30328748330, job 90179708091) that already confirmed both as
  genuine console output belonging to this exact live-tee mechanism.
  5 quotes total (was 3).

Item 31 is now fully resolved -- moved from CLAUDE.md's Active Backlog
to docs/agent-closed-backlog.md's Closed Active Backlog Items section
(keeping its original number), consolidating the full three-PR
resolution history (#423/#424/#425) in one place.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW

* docs: cover both reactive-hint failure conditions in opening sentence

CodeRabbit review: the "Reactive-only failure hint" section documented
two genuine triggers (a real Nuitka compiler failure, and Nuitka
reporting success without producing dist\<env>.exe) but its opening
sentence described only the first. Widened the sentence to cover both
while still excluding the forcefail/HP_TEST_FORCE_NUITKA_FAIL test hooks.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01A4iE1BRSkUETwz237XeuTW

---------

Co-authored-by: Claude <noreply@anthropic.com>