MDView as Markdown

Watcher Capacity

Script Watch, Data Watch, Frame Watch, and Policy Watch all store their observations as a list of unique entries. This list has a finite capacity, and when a watcher grows very large it can reach a point where new observations can no longer be safely recorded.

At Capacity

A watcher is considered at capacity when its list of unique entries has grown large enough that writing new data risks corrupting the existing data. When this happens, the watcher is frozen — no new observations are added, but all previously recorded observations remain intact and can still be viewed.

You will receive an email notification when a watcher reaches capacity.

Broken

In some cases a watcher may already be large when the capacity limit is enforced, and a write may have partially succeeded before the problem was detected. This leaves the watcher in a broken state where the data cannot be read at all.

A broken watcher will show an error on the Inspect page and cannot display any observations until it has been reset.

You will receive an email notification when a watcher becomes broken.

Reducing the Number of Entries

Watchers accumulate entries over time as new unique observations arrive. To reduce the number of unique entries and prevent a watcher from reaching capacity again after a reset, consider the following:

Review your CSP configuration

Each unique blocked resource or script hash generates a separate entry. Tightening your policy to reduce the number of blocked resources will reduce how quickly the watcher grows. If your policy is blocking many resources that should be allowed, updating it to permit them will stop new entries being generated for those resources.

Enable query string removal

If the resources being reported include URLs with unique or rotating query strings (tracking parameters, cache-busting tokens, etc.), each variation is recorded as a separate entry. Enabling query string removal in your Filters will consolidate these into a single entry per path, significantly reducing growth.

Update your site to generate fewer reports

For Script Watch and Policy Watch, consider reviewing which scripts or policies are generating the most reports and whether changes to your site can reduce the volume. Fewer unique reports means fewer watcher entries.

Add noisy URLs to Exclusions (Script Watch, Data Watch, Frame Watch only — not Policy Watch)

If a small number of URLs are responsible for a disproportionate share of new entries (build-hash CDN assets, A/B-test injected scripts, third-party query-string variations, advertising trackers), you can add their URL prefixes to the watcher's Exclusions list. Reports matching an exclusion prefix are skipped at observation time — no entry is recorded, no alert is sent, and the watcher's capacity is preserved.

Exclusions are configured per watcher in the Config modal on the Watch products page. The Config button is on the watcher overview row, next to Inspect, Reset, and Delete.

Each line in the Exclusions list is a literal URL prefix matched against the start of the reported URL. Scheme and host are matched case-insensitively (https://Example.com/ behaves the same as https://example.com/); path and query are matched case-sensitively per RFC 3986. The list has a per-line length cap and a total entry cap to keep the watcher row within its storage budget — invalid lines or lines beyond the caps are dropped on save and a notice is shown.

For example, an exclusion of https://ad.doubleclick.net/activity/ matches every reported URL starting with that prefix (such as https://ad.doubleclick.net/activity/12345 and https://ad.doubleclick.net/activity/N4567890123/) but not https://ad.doubleclick.net/different-path/.

If you have already reset a watcher to clear out noisy entries, adding the relevant prefixes to Exclusions before observations resume will prevent the same entries from re-filling the watcher.

Resetting a Watcher

Resetting a watcher permanently deletes all recorded observations and starts fresh. This is the only way to unfreeze an at-capacity watcher or recover a broken one.

The Reset button is available on the watcher overview page next to each site. After resetting, the watcher will begin recording new observations immediately.

WARNING: Resetting a watcher is permanent. All previously recorded observations will be lost and cannot be recovered.