xmrig-auto-pause: fix stuck state after external restart, add startup cooldown

Two bugs found during live verification on the server:

1. Stuck state after external restart: if something else restarted xmrig
   (e.g. deploy-rs activation) while paused_by_us=True, the script never
   detected this and became permanently stuck — unable to stop xmrig on
   future load because it thought xmrig was already stopped.

   Fix: when paused_by_us=True and busy, check if xmrig is actually
   running. If so, reset paused_by_us=False and re-stop it.

2. Flapping on xmrig restart: RandomX dataset init takes ~3.7s of intense
   non-nice CPU, which the script detected as real workload and immediately
   re-stopped xmrig after every restart, creating a start-stop loop.

   Fix: add STARTUP_COOLDOWN (default 10s) — after starting xmrig, skip
   CPU checks until the cooldown expires.

Both bugs were present in production: the script had been stuck since
Apr 3 (2+ days) with xmrig running unmanaged alongside llama-server.

tests/xmrig-auto-pause.nix

@@ -43,6 +43,7 @@ pkgs.testers.runNixOSTest {
     POLL_INTERVAL = "1"
     GRACE_PERIOD = "5"
     CPU_THRESHOLD = "10"
+    STARTUP_COOLDOWN = "4"
 
     def start_cpu_load(name):
         """Start a non-nice CPU burn as a transient systemd unit."""
@@ -64,6 +65,7 @@ pkgs.testers.runNixOSTest {
             f"--setenv=POLL_INTERVAL={POLL_INTERVAL} "
             f"--setenv=GRACE_PERIOD={GRACE_PERIOD} "
             f"--setenv=CPU_THRESHOLD={CPU_THRESHOLD} "
+            f"--setenv=STARTUP_COOLDOWN={STARTUP_COOLDOWN} "
             f"{PYTHON} {SCRIPT}"
         )
         # Monitor needs two consecutive polls to compute a CPU delta.
@@ -84,8 +86,9 @@ pkgs.testers.runNixOSTest {
 
     with subtest("xmrig resumes after grace period expires"):
         # Already idle since previous subtest. Grace period (5s) plus
-        # detection delay (~2 polls) means xmrig should restart within ~8s.
-        machine.wait_until_succeeds("systemctl is-active xmrig", timeout=15)
+        # detection delay (~2 polls) plus startup cooldown (4s) means
+        # xmrig should restart within ~12s.
+        machine.wait_until_succeeds("systemctl is-active xmrig", timeout=20)
 
     with subtest("Intermittent load does not cause flapping"):
         # First load — stop xmrig
@@ -104,7 +107,7 @@ pkgs.testers.runNixOSTest {
         machine.fail("systemctl is-active xmrig")
 
         stop_cpu_load("cpu-load-2")
-        machine.wait_until_succeeds("systemctl is-active xmrig", timeout=15)
+        machine.wait_until_succeeds("systemctl is-active xmrig", timeout=20)
 
     with subtest("Sustained load keeps xmrig stopped"):
         start_cpu_load("cpu-load-3")
@@ -116,6 +119,20 @@ pkgs.testers.runNixOSTest {
         machine.fail("systemctl is-active xmrig")
 
         stop_cpu_load("cpu-load-3")
-        machine.wait_until_succeeds("systemctl is-active xmrig", timeout=15)
+        machine.wait_until_succeeds("systemctl is-active xmrig", timeout=20)
+
+    with subtest("External restart detected and re-stopped under load"):
+        # Put system under load so auto-pause stops xmrig.
+        start_cpu_load("cpu-load-4")
+        machine.wait_until_fails("systemctl is-active xmrig", timeout=20)
+
+        # Something external starts xmrig while load is active.
+        # The script should detect this and re-stop it.
+        machine.succeed("systemctl start xmrig")
+        machine.succeed("systemctl is-active xmrig")
+        machine.wait_until_fails("systemctl is-active xmrig", timeout=20)
+
+        stop_cpu_load("cpu-load-4")
+        machine.wait_until_succeeds("systemctl is-active xmrig", timeout=20)
   '';
 }