diff --git a/.github/workflows/coyote.yaml b/.github/workflows/coyote.yaml
index c6cd49bf..ca5ade98 100644
--- a/.github/workflows/coyote.yaml
+++ b/.github/workflows/coyote.yaml
@@ -14,8 +14,10 @@ name: Concurrency (Coyote)
 #   - schedule (weekly) + workflow_dispatch — the deep run (10000 iterations/test).
 #
 # The TestKit doubles have their own systematic scenarios (tests/Wolfgang.Etl.TestKit.Tests
-# .Concurrency, #126): xunit-hosted Coyote TestingEngine runs that only execute when the
-# assembly was rewritten and COYOTE_REWRITTEN=1 is set (#365). That project targets net10.0;
+# .Concurrency, #126): xunit-hosted Coyote TestingEngine runs. They always run: the ordinary
+# test sweep runs them un-rewritten in systematic fuzzing mode (timing perturbation, no
+# controlled schedule), and this workflow rewrites the assembly and sets COYOTE_REWRITTEN=1
+# so the same tests run under controlled scheduling (#365). That project targets net10.0;
 # `coyote rewrite` only patches the IL, so the runtime gap above does not apply.
 #
 # NOT implemented: the optional 24h soak (handle/thread/memory-leak hunting via
@@ -132,9 +134,9 @@ jobs:
         run: coyote rewrite rewrite.coyote.json
 
       - name: Run the systematic tests under the rewritten assemblies
-        # RunSystematic early-returns without this flag (the normal PR sweep must not
-        # run un-rewritten Coyote, which reports false deadlocks); with it, each [Fact]
-        # drives a TestingEngine over 500 controlled interleavings and fails on any bug.
+        # Without this flag RunSystematic uses systematic fuzzing mode (what the ordinary
+        # test sweep runs, since un-rewritten Coyote cannot control the schedule); with it,
+        # each [Fact] drives a TestingEngine under controlled scheduling and fails on any bug.
         env:
           COYOTE_REWRITTEN: '1'
         run: dotnet test "$PROJECT/Wolfgang.Etl.TestKit.Tests.Concurrency.csproj" -c Release --no-build --logger "console;verbosity=normal"