diff --git a/.github/workflows/ubuntu-unit-tests.sh b/.github/workflows/ubuntu-unit-tests.sh
index 705b147b96..d48c4b4bfe 100755
--- a/.github/workflows/ubuntu-unit-tests.sh
+++ b/.github/workflows/ubuntu-unit-tests.sh
@@ -47,13 +47,15 @@ for phase in "${PHASES[@]}"; do
             meson test -C build --print-errorlogs
             ;;
         RUN_ASAN_UBSAN|RUN_GCC_ASAN_UBSAN|RUN_CLANG_ASAN_UBSAN)
+            MESON_ARGS=(--optimization=1)
+
             if [[ "$phase" = "RUN_CLANG_ASAN_UBSAN" ]]; then
                 export CC=clang
                 export CXX=clang++
                 # Build fuzzer regression tests only with clang (for now),
                 # see: https://github.com/systemd/systemd/pull/15886#issuecomment-632689604
                 # -Db_lundef=false: See https://github.com/mesonbuild/meson/issues/764
-                MESON_ARGS=(-Db_lundef=false -Dfuzz-tests=true --optimization=1)
+                MESON_ARGS+=(-Db_lundef=false -Dfuzz-tests=true)
             fi
             meson --werror -Dtests=unsafe -Db_sanitize=address,undefined "${MESON_ARGS[@]}" build
             ninja -C build -v
@@ -61,12 +63,17 @@ for phase in "${PHASES[@]}"; do
             export ASAN_OPTIONS=strict_string_checks=1:detect_stack_use_after_return=1:check_initialization_order=1:strict_init_order=1
             # Never remove halt_on_error from UBSAN_OPTIONS. See https://github.com/systemd/systemd/commit/2614d83aa06592aedb.
             export UBSAN_OPTIONS=print_stacktrace=1:print_summary=1:halt_on_error=1
-            # There's some weird stuff going on in GH Actions, where the following
-            # `meson test` command hangs after test #252 unless it's executed under
-            # unbuffer or there's something else producing output. So far it happens
-            # _only_ with ASAn (not with UBSan), both with gcc and clang. I'll
-            # need to take a closer look later...
-            unbuffer meson test --timeout-multiplier=3 -C build --print-errorlogs
+
+            # FIXME
+            # For some strange reason the GH Actions VM stops responding after
+            # executing first ~150 tests, _unless_ there's something producing
+            # output (either running `meson test` in verbose mode, or something
+            # else in background). Despite my efforts so far I haven't been able
+            # to identify the culprit (since the issue is not reproducible
+            # during debugging, wonderful), so let's at least keep a workaround
+            # here to make the builds stable for the time being.
+            (set +x; while :; do echo -ne "\n[WATCHDOG] $(date)\n"; sleep 30; done) &
+            meson test --timeout-multiplier=3 -C build --print-errorlogs
             ;;
         CLEANUP)
             info "Cleanup phase"
diff --git a/src/test/test-execute.c b/src/test/test-execute.c
index 1890cc4f75..83816f474c 100644
--- a/src/test/test-execute.c
+++ b/src/test/test-execute.c
@@ -897,8 +897,8 @@ int main(int argc, char *argv[]) {
         test_setup_logging(LOG_DEBUG);
 
 #if HAS_FEATURE_ADDRESS_SANITIZER
-        if (strstr_ptr(ci_environment(), "travis")) {
-                log_notice("Running on TravisCI under ASan, skipping, see https://github.com/systemd/systemd/issues/10696");
+        if (strstr_ptr(ci_environment(), "travis") || strstr_ptr(ci_environment(), "github-actions")) {
+                log_notice("Running on Travis CI/GH Actions under ASan, skipping, see https://github.com/systemd/systemd/issues/10696");
                 return EXIT_TEST_SKIP;
         }
 #endif