fix(worker): continuous loop + cron as watchdog, not scheduler

The worker is supposed to run back-to-back evals as a continuous daemon,
not wait for a cron trigger. My earlier install had it backwards — the
:07 cron ran one pass then exited, leaving the machine idle between
cron ticks. Inverting:

lem-eval.sh loop now runs 'once' in a tight loop with a 10s
breather between passes (enough to not thrash
logs / cron but short enough to stay busy).
Each 'once' iterates all runnable targets and
commit/pushes per target via run_target.

cron/submit.cron now a watchdog, not a scheduler. Runs at :07
hourly and resurrects the loop if it died.
No-op if the loop is still running.

cron/maintain.cron unchanged — :37 hourly git pull for config.

The 1h watchdog cadence means the worker can be dead for up to 1h
before being restarted. For the current MVP that's acceptable; if we
ever need tighter recovery we can switch to systemd or a shorter cron.

Co-Authored-By: Virgil <virgil@lethean.io>

cron/submit.cron

@@ -1,14 +1,13 @@
-# LEM-Eval submit cron — one pass per hour over all runnable targets.
+# LEM-Eval submit watchdog — ensures the continuous `loop` worker is
+# alive. The real work is done by `lem-eval.sh loop` running back-to-back
+# on each target; this cron just restarts the loop if it died.
 #
 # Runs at :07 past the hour to match the legacy LEM-benchmarks cadence
-# Snider's homelab has been running since the Ollama pipeline days. The
-# flock guard prevents overlap if a previous run is still going (the
-# 26B/31B model targets can take longer than an hour to complete 8-PAC).
+# (back when the Ollama pipeline did hourly git push instead of a
+# continuous loop). The 1h cadence is the maximum gap the machine can
+# be idle if the worker crashes — after that, this cron resurrects it.
 #
 # Install with:
 #   crontab -l | cat - cron/submit.cron | crontab -
-#
-# Adjust the path to the LEM-Eval clone if you installed it elsewhere
-# than $HOME/PycharmProjects/LEM-Eval.
 
-7 * * * * cd $HOME/PycharmProjects/LEM-Eval && flock -n .lock ./lem-eval.sh once >> /tmp/lem-eval.log 2>&1
+7 * * * * cd $HOME/PycharmProjects/LEM-Eval && pgrep -fu "$USER" "lem-eval.sh loop" >/dev/null || nohup ./lem-eval.sh loop >> /tmp/lem-eval.log 2>&1 &

lem-eval.sh

@@ -163,10 +163,15 @@ case "$MODE" in
         maintain
         ;;
     loop)
+        # Continuous worker: run one pass after another with no long sleep.
+        # Each pass iterates every runnable target on this host and commits
+        # + pushes after each target's eval, so the canon on HF stays in
+        # lockstep with the local state. A brief breather prevents a hot
+        # loop if something misconfigures — 10 seconds is enough to not
+        # thrash cron/logs but short enough that the machine stays busy.
         while true; do
             once
-            log "sleeping 1h"
-            sleep 3600
+            sleep 10
         done
         ;;
     *)