P29-T6: TickDynamics re-read + charter — the required re-read and the ServicesProtocol boundary design

[percy-raskova]

ADR198 charter (umbrella #557). TickDynamics @4.0 carries the survey's ONLY required INADEQUATE-COVERAGE re-read: per-scenario dormancy re-derivation across all six county-bearing canonical scenarios (the reader's blanket-dormancy premise was withdrawn — SIX scenarios drive the annual pipeline at tick 52), an explicit RESERVED-LINE section (bifurcation directional score, five-share ClassDistribution/Feature-016, dispossession_cascade_milestones — register rows 21), and a D102-era sweep now superseded by the PR A discharge.

Then the charter: the ~28-field ServicesProtocol boundary design, round() half-even semantics, county-identity encoding per R7 (int-encoded FIPS, leading-zero D-record). Output: a port plan or an honest split verdict.

🤖 Generated with Claude Code

[percy-raskova]

T6 delivered (2026-08-14). The per-scenario dormancy re-read exists as six memos (reports/t6-dormancy/{single_county,mitterrand,syriza,bernie_valve,detroit_tri_county,michigan_canada_e2e}.md) plus the synthesis + ServicesProtocol charter verdict: reports/t6-tickdynamics-services-charter-2026-08-14.md.

Headline — the "tick 52" premise is REFUTED as stated:

• The four qa-harness Wayne scenarios fire the annual pipeline at context tick 0 (the FIRST step() call — pre-increment TickContext, simulation_engine.py:566-569), golden-verified.
• detroit_tri_county's committed artifacts (5t JSON, 5-row dense, 3-tick vault) never fire it — tick 52 is a counterfactual needing --ticks ≥53.
• Only michigan_canada_e2e confirms the premise (9 fires, ticks 52…468, years 2011–2019).
• The withdrawn county-free blanket-dormancy premise (regression_scenarios.py:406-451) applies to none of the six — dormancy is wiring-driven, not substrate-driven.

Charter verdict: ONE ServicesProtocol charter, wiring-class-organized (no split needed): the 33-field surface by step, absent-semantics named per field, defaults-are-behavior (golden-visible), computed zeros pinned as computations, dead-ends port verbatim or retire per row 24's ruling. Director-gated residue: whether the runner's Class-A unwiring (esp. employment_source — flat 100k employment in a scope whose counties differ 2× in population) is deliberate scope-trimming or a data bug. That one wants your answer before the port's golden semantics can be fully pinned.

Full ledger (Class A/B/C wiring, live-but-dead-end channels, the round() half-even ×9 port hazard, the unexercised R7 leading-zero trap, carry-forward asymmetry) in the synthesis report §§2-3.