pp-test: restore drops caches before production load; all-mode EXIT trap auto-restores; ssh ConnectTimeout

scripts/pp-test.sh

@@ -39,8 +39,8 @@ fi
 log()  { echo "[pp-test] $*"; }
 die()  { echo "[pp-test] FAIL: $*" >&2; exit 1; }
 
-ssh_claw()   { ssh -o BatchMode=yes "$CLAW"   "$@"; }
-ssh_server() { ssh -o BatchMode=yes "$SERVER" "$@"; }
+ssh_claw()   { ssh -o BatchMode=yes -o ConnectTimeout=8 "$CLAW"   "$@"; }
+ssh_server() { ssh -o BatchMode=yes -o ConnectTimeout=8 "$SERVER" "$@"; }
 
 # ---------------------------------------------------------------------------
 # preflight
@@ -158,6 +158,9 @@ do_stop_pp() {
 }
 
 do_restore() {
+    # Warm-cache loads crash fast on the claw (AGENTS.md); cold loads survive.
+    log "dropping page caches on claw before production load"
+    ssh_claw "$SUDO sh -c 'sync; echo 3 > /proc/sys/vm/drop_caches'" || log "WARN: cache drop failed"
     log "starting production services"
     ssh_server "$SUDO systemctl start frank-default-small" || die "could not start frank-default-small"
     ssh_claw   "$SUDO systemctl start frank-default-large" || die "could not start frank-default-large"
@@ -185,11 +188,14 @@ case "${1:-}" in
     stop-pp)   do_stop_pp ;;
     restore)   do_restore ;;
     all)
+        log "full run: any failure auto-restores production services"
         do_preflight
+        trap 'do_restore || true' EXIT
         do_start
         do_verify
         do_bench
         do_restore
+        trap - EXIT
         ;;
     *)
         echo "Usage: $0 {preflight|start|verify|bench [prompt_k] [out] [reps]|stop-pp|restore|all}"