4.1.9

Bug Fixes

• Fixed potential over-deletion of log history during the daily cron cleanup. The destructive DELETE was sized using an approximate row count from MySQL metadata, which can drift by orders of magnitude on InnoDB. The cron now gates the delete on an exact byte-size check before deciding what to remove.
• Fixed captured-URL counts undercounting when the same URL was logged with and without leading or trailing slashes (e.g. /foo, foo, /foo/). URL variants now collapse to a single canonical row in the hits rollup, restoring slash-tolerant matching that regressed in 4.1.7.
• Fixed Email Digest dropping captured URLs that had zero recent hits. The "Top Captured 404s" section now lists captured rows with no hits as well as those with hits, matching the on-screen behavior.
• Fixed Email Digest rendering a misleading "No captured 404s in this period" notice while the hits rollup was being rebuilt. The digest now ships with an explicit "Top URLs unavailable: log rollup is being rebuilt" message instead.
• Fixed the high-impact captured-count cache holding a value of zero for 24 hours when the rollup was unavailable or the query errored. The result is now cached only on a successful query, so the count recovers on the next page load.
• Fixed the Stats trend chart cache being polluted with empty results when the trend query errored or timed out.
• Fixed admin search inputs on Page Redirects and Captured 404s rendering a stray "O"-shaped SVG glyph above the search box.
• Fixed the health bar "unavailable" indicator rendering as a transparent dot during rollup rebuild — it now displays as a muted gray dot to communicate the transient state.
• Fixed every frontend 404 falling through to the theme 404 page on installs where the database upgrade could not complete (corrupted DDL files, opcache divergence, restrictive shared hosting). The redirect lookup now falls back to a schema-tolerant query against the existing redirects table, so manual redirects keep firing while the upgrade is stuck. Trade-off: scheduled redirects briefly stop honoring their start/end windows during the degraded window.
• Fixed asynchronous page suggestions failing to surface for long or low-overlap 404 URLs. The async worker now prioritizes recall over worst-case latency, since it already runs out-of-band and is rate-limited.
• Fixed admin tables remaining stuck on "Loading…" indefinitely when the AJAX request to fetch table data hangs (for example, while the plugin is recreating a missing database table on the slow path). The AJAX call now has a 15-second client-side timeout, the retry/fallback path replaces the loading rows with a clear error message instead of just stripping the placeholder attribute, and AJAX errors render a non-blocking admin notice instead of a native browser alert() dialog.

Improvements

• The Captured 404s and Page Redirects admin tables now render quickly even on sites with millions of log rows. The high-impact captured count, the per-row hit population query, the Email Digest top-captured query, and the Stats trend chart now all read from the pre-aggregated logs_hits rollup table, eliminating multi-second full-log scans that could time out behind Cloudflare and other reverse proxies.
• Health bar AJAX is now decoupled from pagination — the redirects table renders immediately while the health bar hydrates in a separate request, so a slow rollup query no longer blocks first paint.
• The daily cron now reads log row counts from MySQL information_schema metadata instead of running a full index scan on every nightly tick.
• Permalink keyword cron updates are now issued as a single bulk SQL query instead of up to 500 single-row UPDATEs per cycle, sharply reducing database load on busy sites.
• The anonymous suggestion-compute endpoint now enforces a per-IP rate limit and rejects unauthorized requests before any plugin classes are loaded.