rdmsm4x-changelog-20260923-1432-rtty-close-out-517-ready-standard-soak-clean

RTTy work closed out: Build 517 remains the App Store candidate with every engineering question answered, the last open performance question is measured and scoped away from the shipping build, and the fleet is whole again.

Scope

Host: rdmsm4x (hub); fleet read + two relaunches. Session claude@rdmsm4x c13f617a, 2026-09-23 14:26 → 14:32 EDT.
App Store: 0.3.842 / Build 517 attached, VALID, PREPARE_FOR_SUBMISSION — unchanged, waiting on Rich.

The question this session answered

ISSUE-20260915-11 (open p1) describes CPU growing with uptime when a dashboard is visible. It was measured on Direct and never scoped, so it read as a risk against the submission. Measured side by side on one host, both dashboards visible, windows not overlapping, 15-minute samples:

over the same 4.0 h	Standard (App Store)	Direct (fleet)
CPU overall	6.9 % of a core	37.5 %
CPU per interval	6.5 % first, 6.8 % median (no trend)	38.7 % first, 42 % median
RSS	216 → 248 max → 213 MB (flat)	508 → 716 MB (+208 MB, ~50 MB/h)

The Standard copy then ran to 26 h 42 m: 8.2 % average CPU, RSS 228 MB vs 222 MB at 16 min — flat over a day. The growth is Direct-only (Application Flows band + ECSNetTop engine, neither built into Standard). No open performance defect applies to Build 517.

Changes (RTTy repo, pushed to origin/backup/fleet, HEAD 52b98ec)

commit	change
`2125db5`	Release evidence §6 (Standard vs Direct soak, with the conclusion) and §7 (fleet state at close-out)
`52b98ec`	SESSION-STATE close-out block

Earlier in this session (2026-09-22, already recorded): 8bd6258 verifier-bundle re-render, 78fa34d library re-pin after review, 6ef9b63 + 2bd593e gate process handling, 6359cd2 resume block.

Verification evidence

Standard-vs-Direct soak log: ~/dev/_handoff/rtty-main-verify-20260922/standard-vs-direct-soak.log (17 intervals + a final reading).
main at bfa4ede (four agent commits landed after my 09-22 gate run): RTTyTests 1129 (8 skipped), PrivilegedCapture 71, LatencyCore 308, 0 failures. Last full 10/10 gate was at 2bd593e; anything after it is ungated.
Fleet: Build 517 on all six since 2026-09-21 19:37, 0 RTTy crash reports in 43 h.

Environment changes on rdmsm4x / fleet

Quit the Standard 517 test copy (the soak instance) after its final reading. The detached sampler finished on its own.
rdmsm4x and rdmbair15m5 had no RTTy process and no crash report (stopped by gate/dev-loop churn, which stops same-bundle instances before launching a staged build). Relaunched and verified: pids 17967 and 8846, Build 517. Undo: quit the app.
No Xcode swap this session; 27.0 is in place. Two non-persistent MTU-1500 routes to Apple remain (17.156.104/24, 17.56.138/24) — re-add after a reboot before any altool upload.

Tickets

Commented ISSUE-20260915-11 with the scoping result (stays open for the fleet build). No new tickets.

Owner actions (Rich only)

Pricing, EU trader status, App Privacy answers (facts section at the top of PRIVACY-ANSWERS.md), Agreements, then Add for Review → Submit. Plus DEC-20260921-02 (re-pin the release toolchain to Xcode 27?) and ISSUE-20260921-20 (should the fleet build keep pinging ping.ui.com?).