diff --git a/src/usage/filter-index.ts b/src/usage/filter-index.ts
index 5b048e30265..72f90161a51 100644
--- a/src/usage/filter-index.ts
+++ b/src/usage/filter-index.ts
@@ -97,7 +97,10 @@ const SQLITE_BUSY_TIMEOUT_MS = 250;
 const RETAINED_WAL_LIMIT_BYTES = 4 * 1024 * 1024;
 const BATCH_BUSY_RETRIES = 3;
 const BATCH_BUSY_RETRY_MS = 5;
-const BUSY_RETRY_DELAYS_MS = [10, 25, 50, 100, 200, 400, 800, 1_600, 3_200] as const;
+// Contention is retried briefly and then degrades to the streaming read, which answers the same
+// request from the ledger. A longer wait only holds a dashboard request (and, through SQLite's
+// synchronous busy timeout, the event loop) behind another writer.
+const BUSY_RETRY_DELAYS_MS = [10, 25, 50] as const;
 // A rebuild-required error resolves as soon as the competing publish settles, so it retries on
 // the original short schedule instead of parking behind the long busy backoff.
 const REBUILD_RETRY_DELAYS_MS = [10, 25, 50] as const;
@@ -1570,9 +1573,9 @@ export async function readUsageFilterCandidates(
         continue;
       }
       if (isBusyError(error)) {
-        // Contention, including a peer still creating the projection, is waited out rather than
-        // read as a foreign file. The bound is the BUSY_RETRY_DELAYS_MS sum (about 6.4 s) plus up
-        // to ten busy timeouts before the request falls back to the streaming read.
+        // Contention, including a peer still creating the projection, is waited out briefly rather
+        // than read as a foreign file. The bound is the BUSY_RETRY_DELAYS_MS sum (85 ms) plus up to
+        // four busy timeouts before the request falls back to the streaming read.
         const delay = BUSY_RETRY_DELAYS_MS[attempt];
         if (delay === undefined) return null;
         testHooks?.beforeBusyRetry?.({ dbPath: path, attempt });
diff --git a/structure/usage-client-attribution.md b/structure/usage-client-attribution.md
index 387b3a1bbf9..5a23a051ab3 100644
--- a/structure/usage-client-attribution.md
+++ b/structure/usage-client-attribution.md
@@ -133,7 +133,8 @@ projection read-only, requires this projection's full table set rather than a co
 table, and only uses it when it is current for the exact ledger revision; otherwise it streams,
 and it never creates, writes or registers anything. The passive read makes a single attempt and
 degrades to the streaming scan on any error; it does not retry. Only the active index read
-(`readUsageFilterCandidates`) retries contention on nine delays totaling about 6.4 s and runs a
+(`readUsageFilterCandidates`) retries contention on three short delays (85 ms, plus SQLite's
+250 ms busy timeout per attempt) before streaming, and runs a
 foreign or unsafe open and a rebuild-required read through three short delays before degrading.
 Any other
 index-side failure -- unwritable home,
diff --git a/tests/usage/usage-filter-index.test.ts b/tests/usage/usage-filter-index.test.ts
index 0ecd44da494..b466567d16a 100644
--- a/tests/usage/usage-filter-index.test.ts
+++ b/tests/usage/usage-filter-index.test.ts
@@ -1345,19 +1345,26 @@ describe("filtered usage index availability", () => {
     appendRows([{ ...syntheticRow(99), requestId: "late", provider: "kimi" }]);
     const lock = new Database(usageFilterIndexPath(testDir));
     lock.exec("BEGIN IMMEDIATE");
+    // The busy back-off is bounded and short: a held lock degrades to the streaming read instead of
+    // holding the request for seconds. Record the scheduled delays rather than timing the run.
+    const busyDelays: number[] = [];
+    setUsageFilterIndexTestHooksForTests({
+      beforeRetryDelay: ({ delayMs }) => { busyDelays.push(delayMs); },
+    });
     try {
       const result = await getFilteredUsageAggregate({ provider: "openai" });
       expect(summary(result).summary.requests)
         .toBe(rows.filter(row => row.provider === "openai").length);
       // Contention is an index problem, not a ledger failure: the request still answers.
       expect(result.filterIndex).toBeUndefined();
+      expect(busyDelays.length).toBeGreaterThan(0);
+      expect(busyDelays.reduce((sum, delay) => sum + delay, 0)).toBeLessThanOrEqual(250);
     } finally {
+      setUsageFilterIndexTestHooksForTests(null);
       lock.exec("ROLLBACK");
       lock.close();
     }
-    // Sustained contention exhausts the extended busy-retry window before degrading to the
-    // streaming read, which takes several seconds by design; the default 5s timeout is too tight.
-  }, 20_000);
+  });
 
   test("an unchanged filtered read releases its index reader", async () => {
     const rows = Array.from({ length: 5 }, (_, index) => syntheticRow(index));