diff --git a/common.sh b/common.sh
index 9dca2b2..ea6c57a 100755
--- a/common.sh
+++ b/common.sh
@@ -2073,6 +2073,48 @@ wd_state_set() {
 # --dry-run / --log are threaded down automatically from $DRY_RUN / $ENABLE_LOGGING —
 # never $VERBOSE, which nothing in this codebase ever assigns.
 
+# Writes the run record for a script an orchestrator ran, in the same shape and the same place
+# run_job.sh writes one for a top-level job: LOG_DIR/<id>.json.
+#
+# Until this existed, only the ten jobs cron starts had a record — every script inside a
+# *_MAINTENANCE_SCRIPTS, WATCHDOG_ORCHESTRATOR_SCRIPTS or ARRAY_*_SCRIPTS list had none, roughly
+# seventy entries. Nothing announced that. Consumers just quietly got less: the arr cleanup stats
+# fell through to a daily aggregate that cannot report an end time, and the troubleshoot profile
+# answered "I cannot see a run record for that" for every script that is not an orchestrator —
+# which is nearly all of them, and exactly the ones a question is usually about.
+#
+# Deliberately NOT run_job.sh. That wrapper also takes a lock, enforces a minimum interval and
+# writes a manual sentinel; running it here would let an orchestrator's own child skip itself for
+# being "too soon" or block on a lock the parent already holds. The record is the part that was
+# missing, so the record is the only part this borrows.
+#
+# Failure is silent by design. A record that cannot be written must never stop the work it was
+# describing — LOG_DIR could be read-only or full, and the script itself still needs to run.
+_orch_child_record() {
+    local rel="${1%.sh}" start="$2" ec="${3-}"
+    [[ -n "$rel" && -n "${LOG_DIR:-}" ]] || return 0
+
+    local f="$LOG_DIR/$rel.json"
+    mkdir -p "$(dirname "$f")" 2>/dev/null || return 0
+
+    if [[ -z "$ec" ]]; then
+        printf '{"id":"%s.sh","status":"running","start":%s,"pid":%s}\n' \
+            "$rel" "$start" "$$" > "$f" 2>/dev/null
+        return 0
+    fi
+
+    # Same ladder run_job.sh uses: 0 ok, 1 warn, anything else error. Kept identical so a record
+    # written here and one written there mean the same thing to everything that reads them.
+    local status
+    if   (( ec == 0 )); then status=ok
+    elif (( ec == 1 )); then status=warn
+    else                     status=error
+    fi
+    printf '{"id":"%s.sh","status":"%s","start":%s,"end":%s,"exit":%s}\n' \
+        "$rel" "$status" "$start" "$(date +%s)" "$ec" > "$f" 2>/dev/null
+    return 0
+}
+
 run_orch_child() {
     local entry="$1"
     local script_args script_path script_name label extra_args run_args
@@ -2098,12 +2140,15 @@ run_orch_child() {
     local _start _ec
     _start=$(date +%s)
     log "Running: $label"
+    _orch_child_record "${script_args[0]}" "$_start"
     if bash "$script_path" "${run_args[@]}"; then
         log "$script_name — done in $(format_duration $(( $(date +%s) - _start )))"
+        _orch_child_record "${script_args[0]}" "$_start" 0
         JOB_PASS+=("$label")
         return 0
     else
         _ec=$?
+        _orch_child_record "${script_args[0]}" "$_start" "$_ec"
         # Match the severity ladder run_job.sh already records against: exit 1 is a warning,
         # exit 2+ is an error. The status in the run record was always derived that way, so a
         # child exiting 1 produced a run marked "warn" whose log was full of ❌ ERROR lines —