gHashTag · Oct 17, 2026 · Oct 17, 2026
@@ -0,0 +1,41 @@
+{
+  "built_by": "t27c-bootstrap@0.5.2+cc80a305d",
+  "closure": [
+    {
+      "spec": "specs/queen/control.t27",
+      "spec_hash": "sha256:2c9c2698bca4bd146d21cf8ee002935e8c80cb88a0fc29cc5bb75e90eb496e58"
+    },
+    {
+      "spec": "specs/queen/keyed.t27",
+      "spec_hash": "sha256:2c396c47a4b27d995c79e8a1b91ad87d3772463aa9a6d7794e78aed30a4c8b3e"
+    },
+    {
+      "spec": "specs/queen/keyed_guard.t27",
+      "spec_hash": "sha256:b76136cb283f2374d681b77316138d96fddeb3fd1bff5316879461ba79a409f8"
+    }
+  ],
+  "config": "gen=c,rust,verilog,zig;tests=zig",
+  "gen_hash_c": "sha256:016a3b5765e38a6d5ee1ec3f580bea1f1a62cbc2998ec40096f3072b9c28d3d5",
+  "gen_hash_rust": "sha256:cd10d95cea73ac75ebbde491d64f36c9c722bcc1ae2e9cdd596c9166db50bc6f",
+  "gen_hash_verilog": "sha256:0999747556d0fb5cf5e01c582e7de6eeb917fd053ac38c994e5055fee3afeb9c",
+  "gen_hash_zig": "sha256:3432ae04747785e4e3c63470422070731c3e3920b381d30d2577793682528fb4",
+  "module": "QueenKeyedBatch",
+  "ring": 12,
+  "seal_schema": 2,
+  "sealed_at": "2026-10-17T17:26:49Z",
+  "sealed_by": "t27c-bootstrap@0.5.2",
+  "spec_hash": "sha256:80709724eab9fa180050984b0586ef224c87f6a8b70db245d3d04bb5af177006",
+  "spec_path": "specs/queen/keyed_batch.t27",
+  "tests": {
+    "failed": 0,
+    "failing": [],
+    "forced": false,
+    "passed": 13,
+    "total": 13,
+    "vacuous": 0
+  },
+  "toolchain": {
+    "test_runner": "sha256:1982ae948e24d00c5d221ccec12dfce1eb01204b53e37b0f857f93b16d64629d",
+    "zig": "0.16.0"
+  }
+}
@@ -0,0 +1,23 @@
+# NOW -- keyed_batch.t27: the bee admission admits a batch of ready issues per queend choose (2026-10-18)
+
+## specs/queen/keyed_batch.t27 (Closes #10033; epic #7851)
+
+- Why: the actors trial of 2026-10-17 (#7851, comment 6237225782) started 2 bees in 20 minutes, against the loop's 18.5. The round readies about 826 candidates. The one admission actor then asked queend once per issue (`await deps.admit`, `queen-bee-actors.ts:298` at trios 42042a5), while the loop asks once per round.
+- What: the admission now groups asks into one queend choose. keyed_guard.t27's answers stay the same; only the number of queend calls changes.
+  - `joins_batch` and `answer_now`: exactly the asks `admit_answer` would send to queend join a batch, and the rest are answered at once.
+  - `batch_goes_out`, `batch_take` and `chooses_for` (group commit): a batch goes out when no choose is in flight, with no timer, up to `BATCH_ASKS_MAX` = 1024, which is at least 826.
+  - `new_lane_quota`: control.t27 `batch_quota`, the loop's cap.
+  - `word_of`, `batch_answer` and `batch_grants`: queend's answer gives each issue PICK, LATER or NO. A lone ask answers as `admit_answer` does, and a batch grants exactly its quota.
+  - `batch_wake_count`: a freed lane wakes every waiting issue into one batch.
+  - `round_chooses` (the model): the measured round of 826 candidates on 47 lanes, with one pick in ten, takes 461 serial chooses and 1 batched.
+  - `ask_wait_ms`, `ask_bound_ok` and `word_after_choose` (bounds): an ask's call must outlast two chooses. `askQueend` has no bound today. A choose that ends with no answer moves no lane.
+- Lab evidence (Railway lab, master 9a7d54aaf tree, t27c 0.5.2):
+  - `t27c test-report`: 13 of 13 tests pass, 0 vacuous, and 6 invariants are proved at comptime.
+  - `tri mutate spec`: 85 of 89 mutants are killed. Of the other four:
+    - 2 survivors are equivalent (`<` to `<=` where both branches return `BATCH_ASKS_MAX`);
+    - 1 is a hang from the mutant's own endless loop;
+    - 1 is unviable, because an invariant rejects it at comptime.
+  - Hand constant plants: 10 of 10 are caught.
+  - Wasm build (gen-c plus zig cc wasm32-wasi): reproducible, sha256 `1ed6bcb1...cfd8a`, 1187 bytes.
+  - Seal: verified, all hashes match.
+- Runtime: the thin TypeScript wiring is a gHashTag/trios issue. It covers a batch-shaped `deps.admit`, `planBatch` over one choose, `batch_wake_count`, and a bound on `askQueend`. The before/after numbers go on #7851.