diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 589c0bb..2affde2 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -112,7 +112,7 @@ jobs:
           - {suite: browser-integration, shard: 2, shards: 3}
           - {suite: browser-integration, shard: 3, shards: 3}
     runs-on: ubuntu-latest
-    timeout-minutes: 15
+    timeout-minutes: ${{ matrix.suite == 'browser-conformance' && 20 || 15 }}
     env:
       UV_PYTHON: '3.14.7'
     steps:
diff --git a/docs/interpreter-exceptions.md b/docs/interpreter-exceptions.md
index 61b5aff..b8246a7 100644
--- a/docs/interpreter-exceptions.md
+++ b/docs/interpreter-exceptions.md
@@ -566,3 +566,11 @@ count divided by ten, matching the shared validation inventory's cost basis.
 Small selections retain the existing bound; semantic execution, work/allocation
 limits and timeout/failure diagnostics are unchanged. This addresses the two
 fixed-deadline Linux CI failures in the first batch's stepped/parked matrix.
+
+LA-58B retains the complete browser matrix with a finite 20-minute conformance
+job bound; integration jobs keep their 15-minute limit. Two 15-minute CI
+attempts reached the final compatibility check after their preceding checks
+passed. The identical complete 44-check shard passes locally in 485.824 seconds.
+All six conformance and three integration shards, reports and semantic/resource
+limits are retained; the deterministic partition test uses the actual six
+conformance shards. Interpreter/package/offline acceptance evidence is reused.
diff --git a/tests/test_test_runner.py b/tests/test_test_runner.py
index a0f964b..29a8307 100644
--- a/tests/test_test_runner.py
+++ b/tests/test_test_runner.py
@@ -18,7 +18,7 @@ def test_focused_selection_contains_only_matching_node_checks(self):
         self.assertEqual({check.suite for check in selected}, {'node-unit', 'node-conformance'})
 
     def test_shards_are_complete_disjoint_balanced_and_deterministic(self):
-        for suite, count in [('interpreter', 4), ('browser-conformance', 4), ('browser-integration', 3)]:
+        for suite, count in [('interpreter', 4), ('browser-conformance', 6), ('browser-integration', 3)]:
             with self.subTest(suite=suite):
                 selected = run_tests.select_checks([suite])
                 groups = run_tests.shards(selected, count)