test: replace final fixed-delay oracles

docs/reviews/2026-08-18-r3-test-truth-ledger.md

@@ -53,8 +53,8 @@ inventory refreshed for Batch L reports:
 | Constant-truth assertion methods / sites | 0 / 0 |
 | Reviewed syntactic self-comparison methods / sites | 4 / 5 |
 | Reviewed diagnostic methods | 57 |
-| Methods directly using `Thread.Sleep` / `Task.Delay` | 15 / 17 |
-| Cancellable-infinite-only / elapsed-time methods | 5 / 27 |
+| Methods directly using `Thread.Sleep` / `Task.Delay` | 15 / 14 |
+| Cancellable-infinite-only / elapsed-time methods | 5 / 24 |
 | Methods directly reading environment variables | 47 |
 | Methods directly reading `.cs` source text | 63 |
 
@@ -147,7 +147,7 @@ identifiable in the lane report.
 | T-016 historical test taxonomy | first descriptive-identity batch active in batch J | Remove opaque AP/R/J/K/Slice codes from current test/type names while preserving provenance in comments and ledgers. The 47 `Issue###` files remain an explicit user-decision set because their IDs still connect tests to retail evidence. |
 | T-017 Avalonia ownership | reopened and closed in batch D | The full gate exposed the same compositor ownership class between six newer `MainWindowViewTests` facts. Their six named assertion phases now run in one owned Avalonia application session; 25 fresh-process stress iterations and the complete gate pass. |
 | T-018 stderr observer race | resolved in R2 | Live reader shares writes/deletes; 25 focused repetitions plus the complete gate. |
-| T-019 remaining explicit waits | classified in batch H; six fixed-delay negative oracles resolved in batches M/N, three remain | Five methods are deterministic cancellation stubs, not wall-clock tests. Preserve intentional real-time contracts and bounded integration polling. The remaining cleanup owners are headless late-subscriber replay and two launcher installer/recovery contracts. |
+| T-019 remaining explicit waits | resolved in batches M–O | All nine fixed-delay negative oracles now use observed monitor/thread/lease/process state. The remaining 24 elapsed-time methods are the reviewed cancellation, real-time protocol, cooperative-yield, bounded-polling, and positive-timeout categories from Batch H. |
 | T-020 exact duplicate bodies/data | resolved in batch K | Four redundant executions were removed. The remaining 11 body-equivalent groups are intentionally split theories with disjoint, meaningfully named datasets; the inventory enforces zero repeated rows within a theory or across body-equivalent theories. |
 | T-021 suspicious assertion shapes | resolved in batch L | The inventory enforces zero `Assert.True(true)` / `Assert.False(false)` sites and reports syntactic self-comparisons for review. All five current self-comparisons assert meaningful determinism or stable identity. |
 
@@ -448,7 +448,7 @@ invocation instead of treating both method-level booleans as equivalent. The
 | Intentional real-time protocol/timeout contract | 4 | Keep with explicit bounds: two Live handshake race delays, the one-second net-probe cadence, and continuous unrelated shutdown drain. |
 | Cooperative yield while virtual-clock/background transport work drains | 8 | Retain for now; prefer an observable receiver/worker signal when that seam exists. The virtual behavioral oracle does not derive from the sleep duration. |
 | Bounded completion/readiness polling | 11 | Retain as integration polling with a terminal assertion and deadline; improve opportunistically with events, not by busy-spinning. |
-| Fixed delay used to prove another operation is still blocked | 9 originally; 3 remain | Batches M/N replace six with observed thread wait state. Replace the remaining elapsed-time absence-of-completion checks with an observed wait state or test seam. |
+| Fixed delay used to prove another operation is still blocked | 9 originally; 0 remain | Batches M–O replace all nine with observed monitor/thread/lease/process state. |
 | Positive completion timeout guard | 1 | Keep. The two-second `WhenAny` in `RuntimeCharacterStateTests` fails only if the operation does not complete; it does not delay a passing run. |
 
 The original nine fixed-delay negative oracles are:
@@ -696,3 +696,41 @@ Batch N verification:
 
 Three fixed-delay negative oracles remain: headless concurrent late-subscriber
 replay and the two launcher installer/recovery contracts.
+
+## Batch O final fixed-delay negative oracles
+
+Batch O removes the last three timing-based absence-of-completion tests:
+
+- `HeadlessPluginSessionTests.LateSubscriberReplayQueuesConcurrentRegistrationExactlyOnceInOrder`
+  now runs registration on a dedicated thread and observes it enter
+  `WaitSleepJoin` while Runtime's exact replay read lease is held;
+- `LauncherInstallerTests.IndependentInstallersSerializeAndWaitingCancellationTouchesNothing`
+  now waits until the second installer has actually encountered the held
+  transaction lease before cancelling it; and
+- both rows of
+  `LauncherInstallerTests.OrphanBakeCanNeverPublishAfterRestartRecovery` now
+  observe publication-lock contention where applicable and wait for the exact
+  orphan child process to exit before checking that it cannot publish later.
+
+The launcher observation points are internal callbacks on the existing
+transaction/publication lease retry paths. They are unset in production, add no
+new public API, and do not change lock, retry, cancellation, or publication
+behavior. The former 150/200 ms delays and post-exit settle delay are gone.
+
+Batch O verification:
+
+- the one focused Headless fact and three focused Launcher cases pass;
+- 25 fresh-process Headless iterations and 25 fresh-process Launcher iterations
+  pass, totaling 100/100 exercised cases;
+- the refreshed inventory reduces direct `Task.Delay` methods from 17 to 14
+  and elapsed-time methods from 27 to 24;
+- none of Batch H's nine fixed-delay negative-oracle methods contains a direct
+  sleep or delay;
+- the complete 44-project Release build reports zero warnings and zero errors;
+  and
+- the no-retry complete hermetic Release gate remains 14,382/14,382 with zero
+  skips or failures across all 12 test assemblies.
+
+The 24 remaining elapsed-time methods retain their reviewed Batch H
+classification: cancellation fakes, intentional real-time protocol behavior,
+cooperative yields, bounded integration polling, or a positive timeout guard.