diff --git a/docs/timing-design.md b/docs/timing-design.md
index 28837028..e4a64a4f 100644
--- a/docs/timing-design.md
+++ b/docs/timing-design.md
@@ -539,12 +539,21 @@ runs, not the cloud's; the cloud's container was 4 cores, the bench is 8.
     delivery, as designed: 84 callbacks a second through `sleep_ms` with the
     main thread idle, 0 while it spins, and `report()` answers over the serial
     console. Same mechanism as the unix build in Phase 4.
-  - **Pending:** the LVGL launcher and drum machine on the phone (kept the
-    screen at brightness 1 for photosensitivity and stayed within the P4/phone
-    windows); the mechanism they would exercise, main-thread bytecode
-    delivery, is what the Android numbers already prove. PyScript/Pyodide
-    pages (the `asyncio` source, the same code the Jupyter proof runs) and an
-    `mp-wasm` rebuild carrying the bridge fix remain the two browser follow-ups.
+  - **Follow-ups (2026-09-26, after 0.6.1).** The browser checks are done.
+    Under PyScript/Pyodide, `lv_test_timer` shows `Timer: asyncio/idle` and
+    its seconds count and arc keep moving. An `mp-wasm` rebuild carrying the
+    bridge fix passes `wasm_host.mjs` under node, runs `lv_test_timer` in the
+    direct gallery host (`Timer: wasm/idle`), brings all 20 portal heroes
+    up with no errors, and passes a new workbench timers test. The runtime
+    that shipped with 0.6.1 threw `function signature mismatch` on every
+    timer callback in all three places. The vendored runtimes are in
+    PyDevices.github.io#9 and workbench#1, and the runner's
+    `MULTIMER_BACKEND` line is dropped with its recipes moved to 0.6.1 in
+    android-runner#24.
+    **Still pending:** the LVGL launcher and drum machine visual pass on the
+    phone. An APK from android-runner#24 builds and installs, but the S21's
+    secure lock screen kept the activity from ever drawing, so nothing was
+    seen. The desktop launcher on 0.6.1 was checked under Xvfb instead.
 - **Phase 6, the deliverables (here).** The four repository series were
   exported with `git format-patch` from branches on each repository's
   `origin/main`, then re-applied with `git am` onto a fresh checkout of each
diff --git a/docs/timing-hardware-tests.md b/docs/timing-hardware-tests.md
index f8613e35..13102d20 100644
--- a/docs/timing-hardware-tests.md
+++ b/docs/timing-hardware-tests.md
@@ -218,5 +218,9 @@ per-plan record of pass/fail and anything found.
   mechanism it exercises is what the numbers already prove. `--install-apk`
   installs the 0.2.3 release; a prior local-key debug build must be uninstalled
   first.
-- **Not run:** PyScript/Pyodide in a browser (the `asyncio` source, proven via
-  Jupyter in the cloud) and an `mp-wasm` rebuild carrying the bridge fix.
+- **Browsers, 2026-09-26 — pass.** PyScript/Pyodide: `lv_test_timer` reports
+  `Timer: asyncio/idle` and animates. The `mp-wasm` rebuild with the bridge fix
+  runs the same demo in the direct gallery host (`Timer: wasm/idle`, 298 → 495
+  deliveries in a second), all 20 portal heroes, and workbench's simulator. The
+  runtime that shipped with 0.6.1 threw `function signature mismatch` on every
+  timer callback. Details: [timing-design.md](timing-design.md#ledger).
diff --git a/lib/multimer/_hostloop.py b/lib/multimer/_hostloop.py
index 3db7205b..dec03ba1 100644
--- a/lib/multimer/_hostloop.py
+++ b/lib/multimer/_hostloop.py
@@ -94,6 +94,10 @@ def _cmdline_tokens():
             return _win32_cmdline()
         except Exception:
             pass
+    # CPython 3.10+ keeps the flags itself (macOS has no /proc).
+    argv = getattr(sys, "orig_argv", None)
+    if argv:
+        return tuple(argv)
     return ()
 
 
@@ -183,7 +187,9 @@ def interactive():
         flags = getattr(sys, "flags", None)
         if getattr(flags, "interactive", 0):
             return True
-        return _main_file() is None
+        # ``python -c`` has no ``__main__.__file__`` either, but no prompt
+        # follows it: it is a batch entry, like ``-m``.
+        return _main_file() is None and not batch()
     if _mcu() and _impl() == "circuitpython":
         # CircuitPython cannot tell code.py from the REPL, and delivers
         # nothing in the background either way: the exit hook is the only
diff --git a/tests/test_hostloop.py b/tests/test_hostloop.py
index fb589b2d..3d6e7785 100644
--- a/tests/test_hostloop.py
+++ b/tests/test_hostloop.py
@@ -121,6 +121,19 @@ def test_circuitpython_firmware_is_never_interactive(self):
         self._fake_impl("circuitpython", "rp2")
         self.assertFalse(_hostloop.interactive())
 
+    def test_cpython_dash_c_is_not_interactive(self):
+        """``python -c`` has no ``__main__.__file__``, and no prompt follows it."""
+        self._patch(_hostloop, "_impl", lambda: "cpython")
+        self._patch(_hostloop, "_main_file", lambda: None)
+        self._patch(_hostloop, "_cmdline_tokens", lambda: ("python", "-c", "import app"))
+        self.assertFalse(_hostloop.interactive())
+
+    def test_cpython_bare_prompt_is_interactive(self):
+        self._patch(_hostloop, "_impl", lambda: "cpython")
+        self._patch(_hostloop, "_main_file", lambda: None)
+        self._patch(_hostloop, "_cmdline_tokens", lambda: ("python",))
+        self.assertTrue(_hostloop.interactive())
+
     def test_circuitpython_firmware_takes_the_exit_hook(self):
         self._fake_impl("circuitpython", "rp2")
         self._patch(_hostloop, "_cmdline_tokens", lambda: ())