watchdog: make WAITING-UNTIL work for footer_ui backends + cap runaway defers

_parse_waiting_until scanned only the pane's last non-empty line for footer_ui backends (claude/
opencode) — but their input-box footer always renders BELOW the agent's final message, so the marker
was never seen and WAITING-UNTIL was effectively dead for claude agents. It's only consulted once the
pane is already idle, so scan the whole capture and take the most-recent marker (the footer never
contains it). Add waiting_until_max (default 7200s) so an agent can't park its own reboot forever.
Tests: footer-honors-marker-above-footer, takes-most-recent, defer + cap in stall_check_one; make the
stall harness's patch() idempotent so a re-patched name doesn't leak into tearDown. 66 pass.

agents.py

@@ -541,17 +541,16 @@ def _last_nonempty_line(text):
     return ""
 
 def _parse_waiting_until(cfg, agent, pane):
-    if backend_of(cfg, agent).get("footer_ui"):
-        line = _last_nonempty_line(pane)
-        if not line.startswith("WAITING-UNTIL:"):
-            return None
-        m = re.search(r"WAITING-UNTIL:\s*(\S+)", line)
-    else:
-        m = re.search(r"WAITING-UNTIL:\s*(\S+)", pane)
-    if not m:
+    # Only consulted once the pane is already idle (see stall_check_one), so scanning the whole
+    # capture and taking the MOST-RECENT marker is safe — and it's the only thing that works for a
+    # footer_ui backend (claude/opencode), whose input-box footer always renders BELOW the agent's
+    # final message. The footer never contains the marker, so the last match is the agent's own
+    # signal, whether or not a status footer follows it.
+    matches = re.findall(r"WAITING-UNTIL:\s*(\S+)", pane)
+    if not matches:
         return None
     try:
-        return datetime.fromisoformat(m.group(1).replace("Z", "+00:00")).timestamp()
+        return datetime.fromisoformat(matches[-1].replace("Z", "+00:00")).timestamp()
     except Exception:
         return None
 
@@ -578,9 +577,19 @@ def stall_check_one(cfg, agent):
     grace = int(cfg["watchdog"].get("stall_grace", 180))
     until = _parse_waiting_until(cfg, agent, pane)
     if until is not None:
-        if now <= until + grace:
+        # An agent that starts a long remote/async run (remote cargo build, terraform apply, VM
+        # provision, long ssh) prints `WAITING-UNTIL: <ISO8601>` so the watchdog holds off instead
+        # of killing it mid-run. Cap how far out it can push its own reboot, so a runaway can't park
+        # itself forever ("some max no matter what").
+        wu_max = int(cfg["watchdog"].get("waiting_until_max", 7200))
+        if wu_max and idle > wu_max:
+            # idle here ≈ how long the pane has sat quiet since the agent emitted the marker; once
+            # that exceeds the cap we reboot no matter how far out the stated deadline is.
+            reason = f"WAITING-UNTIL exceeded the {wu_max}s cap (idle {int(idle)}s) — rebooting regardless"
+        elif now <= until + grace:
             return
-        reason = f"past its WAITING-UNTIL by {int(now-until)}s — self-wake did not fire"
+        else:
+            reason = f"past its WAITING-UNTIL by {int(now-until)}s — self-wake did not fire"
     else:
         stall_idle = int(backend_of(cfg, agent).get("stall_idle", 300))
         if idle < stall_idle: