Guard arr cache writes against in-flight rescans, add rescan monitor tool

A direct arr_cache_write() call mid-rescan wrote a partial snapshot that
looked like real data loss to every consumer of the cache. The guard now
lives in arr_cache_write() itself so every caller is protected, not just
arr_get_tracked_data(). arr_rescan_monitor.sh closes the resulting gap for
rescans triggered outside arr_full_rescan.sh's own trigger-and-wait path.

Tools/Manual-Tools.md

@@ -22,6 +22,7 @@ making any changes.
 - [arr_profile_enforcer.sh](#arr_profile_enforcersh)
 - [webhook_setup.sh](#webhook_setupsh)
 - [ramdisk_stop.sh](#ramdisk_stopsh)
+- [arr_rescan_monitor.sh](#arr_rescan_monitorsh)
 - [Adding a New Tool](#adding-a-new-tool)
 
 ---
@@ -733,6 +734,60 @@ ramdisk_stop.sh --log        # verbose — show each step
 
 ---
 
+## arr_rescan_monitor.sh
+
+Waits for an already-active rescan-type command (RescanFolders/RescanSeries/RescanMovie,
+DownloadedXScan, RefreshX) on the given arr to finish, then refreshes the shared tracked-data
+cache (`arr_cache_write()` in common.sh) with the real post-scan numbers. Triggers nothing
+itself — for watching a rescan that was already started some other way (manual API call,
+the arr's own UI, an ad-hoc debugging session).
+
+### When to Use
+
+```
+You manually trigger a RescanFolders/RescanSeries/RescanMovie outside arr_full_rescan.sh
+  → arr_cache_write() correctly refuses to write while it's active (2026-07-17 — a mid-scan
+    write once read as real data loss to every consumer of the cache)
+  → nothing writes the real number through once the scan finishes unless something is
+    watching for completion — that's this tool
+
+arr_full_rescan.sh's weekly run finds one arr already mid-rescan and skips it
+  → that arr's cache stays stale until its own rescan finishes; run this to close the gap
+    without waiting for next week's scheduled run
+```
+
+### What It Does NOT Do
+
+Does not trigger a rescan — if none is active, it exits immediately with nothing to wait
+for. Use the arr's own UI/API, or `Arrs_Stack/arr_full_rescan.sh`, to start one.
+
+### Usage
+
+```bash
+# Wait for lidarr's active rescan (if any) to finish, then refresh its cache:
+arr_rescan_monitor.sh lidarr
+
+# Override the default 2-hour wait ceiling (seconds):
+arr_rescan_monitor.sh lidarr 3600
+
+# Check whether a rescan is active and the cache's current age, no waiting:
+arr_rescan_monitor.sh lidarr --status
+
+# Show what would be waited on without waiting or writing the cache:
+arr_rescan_monitor.sh lidarr --dry-run
+```
+
+### Notes
+
+- Safe to run against sonarr/radarr the same way — arr_type is the only required arg
+- If the wait times out (default 7200s), the cache is left untouched — re-run once the scan
+  actually finishes, or let it pick up naturally next time something calls
+  `arr_get_tracked_data()` and finds the cache stale with no active rescan
+- Does not conflict with `arr_full_rescan.sh` — that script already refreshes its own cache
+  on completion for scans it triggers itself; this covers everything it didn't start
+
+---
+
 ## docker_prune_images.sh
 
 Removes orphaned Docker images that accumulate after container updates. Two modes: