archilyzer

commit 6a469e4e7ac5e4d8ed03d0a6a39aae7fa4543f8b
parent 93be9ca0422fd29d023614b1d88529127fd5c1e7
Author: I Mean I'm Just Saying <imeanimjustsaying@kiwifarms.st>
Date:   Wed, 30 Sep 2026 00:34:07 -0400

editor: re-review L11 — the health pass's block sits above the storage boot probe's comment, which no longer calls itself the only thing an idle boot runs

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

Diffstat:
M
editor/instrumentation.ts
 | 
30
+++++++++++++++---------------

1 file changed, 15 insertions(+), 15 deletions(-)
diff --git a/editor/instrumentation.ts b/editor/instrumentation.ts
@@ -85,21 +85,6 @@ export async function register() {
     /* a failed re-pause must not block server readiness */
   }
 
-  // ONE PASS OVER THE STORAGE LOCATIONS, and it runs on an IDLE BOOT TOO —
-  // the only thing below the reaper that does.
-  //
-  // The pass probes each location (is the disk here, and if not, where?) and,
-  // for a location the operator armed with `autoRepoint`, re-points it to
-  // wherever the volume actually came up. Probing is read-only — one findmnt
-  // per location — and it is exactly what a machine that has just rebooted with
-  // its platter on a different mountpoint needs done before anybody opens a
-  // page. What idle boot refuses is STARTING WORK, so `enqueue: !idle` is the
-  // whole of the gate: an idle container learns where its disks are and
-  // enqueues nothing.
-  //
-  // Lazy-imported and `void`ed like everything else here: the controller pulls
-  // in execa transitively (Node-only), and a disk that cannot be probed must
-  // never block server readiness.
   // THE DRIVE HEALTH PASS, every 15 s — and ON AN IDLE BOOT TOO, like the
   // probe below. It reads each location's block device counters (never the
   // drive) and keeps, in memory only, which drives are not answering; every
@@ -117,6 +102,21 @@ export async function register() {
     /* a health pass that fails to arm must not block server readiness */
   }
 
+  // ONE PASS OVER THE STORAGE LOCATIONS, and it runs on an IDLE BOOT TOO: it
+  // only reads, like the health pass above.
+  //
+  // The pass probes each location (is the disk here, and if not, where?) and,
+  // for a location the operator armed with `autoRepoint`, re-points it to
+  // wherever the volume actually came up. Probing is read-only — one findmnt
+  // per location — and it is exactly what a machine that has just rebooted with
+  // its platter on a different mountpoint needs done before anybody opens a
+  // page. What idle boot refuses is STARTING WORK, so `enqueue: !idle` is the
+  // whole of the gate: an idle container learns where its disks are and
+  // enqueues nothing.
+  //
+  // Lazy-imported and `void`ed like everything else here: the controller pulls
+  // in execa transitively (Node-only), and a disk that cannot be probed must
+  // never block server readiness.
   // Kept so the queued-meta pass below can wait for it: a re-queue for a
   // channel whose location is mid-autoRepoint would be refused as unreachable.
   let storagePass: Promise<unknown> = Promise.resolve();