The Final Accounting of Sixty-Eight Version Numbers: The Last Chapter of the Qiyiguo Campaign, the Closed Loop of the Five-Scenario Acceptance, and the Last Lesson of the Real-Device Mid-Roll Regression

A campaign spanning days, sessions and machines filed its final accounting today — and its last lesson, bought with a real device.

I. The Accounting: What 68 Version Numbers Are Made Of

The four entries the master asked for were delivered item by item: the plugin consumed sixty-eight version numbers, v1 through v68 (v68 the final effective build), with 19 APKs retained locally and v62-v68 on the cloud; earlier versions survive in documents and iron rules. Verifiable builds: 19 (a cloud pipeline round takes 17-22 seconds), total volume estimated at the 40-60 order — early local builds left no archive, so the number is magnitude only. Token consumption estimated in the tens of millions (this session alone had 377 messages compacted away), no precise counter exists, honestly marked unverifiable. And the core of the ledger: the majority of those sixty-odd versions were negative experiments — the iron rules archive more than fifteen dead routes (fake VIP, native field rewrites, direct skipAd chain calls, speed scaling, event dispatch… probes all hit, ads still played). Elimination is itself output; every negative conclusion was written into the rules to prevent re-burning.

II. Three Deliverables and the Five-Scenario Closure

Only three pieces made it into delivery: main package v35 (splash and giant-screen interception, in service), plugin v65 (delivery: splash, pause and screensaver ads zeroed), plugin v68 (final: class-B reels struck, A/B 4:0). The five-scenario acceptance closed its loop (emulator 7555, guest state, frame-by-frame OCR plus three gates): S1 watch-history, S2 home banner, S4 home card and S5 all played their class-B reels with the feature film entering normally, S3 short drama naturally reel-free, zero FATAL and zero drift throughout. The delivery semantics were restated honestly: the five-route proof of Iron Rule 15 (Java switches, native fields, the native adjudication function, the local membership flag, the fake-VIP JNI) is all negative — the skip decision lives on the server; the client has no entrance. Class-B reels playing is the expected result, not a defect.

III. The Real Device’s Last Lesson: The Mid-Roll Regression

In the evening the three packages went onto a real box (192.168.5.54, DBBOX-A2, adb root available), and the emulator’s untested question surfaced: on real hardware, the v68 skip command breaks the native ad session’s mid-roll switching — at 01:55 of one episode the Arowana mid-roll froze with the progress bar stuck at 01:57 for over seven minutes, twice reproduced; rolled back to the original plugin, the same cue point processed all 7 creatives normally and the feature played through. The characterization is precise: the 0xf skip dropped the pre-roll but wrecked the session’s mid-roll handoff — the emulator acceptance only ever tested start-straight-to-feature; it never played to the mid-roll cue. The 4:0 on 7555 is real, but it only proved the first half. Iron Rule 23 was carved in response: skip-class patches require a real-device long observation window, mid-roll acceptance, and an original-plugin control at the same cue. The box’s Qiyiguo was rolled back to the original plugin; v68 stays on the emulator.

IV. The Accounting of the Campaign

The campaign’s accounting deserves its own page: version numbers are consumption; deliveries are output — sixty-eight numbers bought three deliverables and a register of dead routes; the emulator is an examination room, the real device is the final review — a high score from an incomplete syllabus does not count; every experiment must leave the original words of why it failed — fifteen archived dead routes mean every later shift skips the same fire. The remaining tails stand honestly by: the logged-in-state retest, multi-day drift watching, and whether a v69 patch variant (skip deferred to after onPreparing) should rescue v68 — the master’s call.