WordPress makes several calls to wp_cache_delete() (especially wp-includes/meta.php ): if one of them ends up being a no-op, the relevant cache entry becomes stale. WordPress eventually will re-connect to Redis, and that will respond with a cache hit instead of a cache miss.
Note it is borderline impossible to replicate this behaviour unless the Redis server is hosted on a different machine rather than the one where WordPress runs (with php-fpm, in my case).
I have one issue with
WP_REDIS_GRACEFUL
WordPress makes several calls to
wp_cache_delete()
(especially wp-includes/meta.php ): if one of them ends up being a no-op, the relevant cache entry becomes stale. WordPress eventually will re-connect to Redis, and that will respond with a cache hit instead of a cache miss.Note it is borderline impossible to replicate this behaviour unless the Redis server is hosted on a different machine rather than the one where WordPress runs (with php-fpm, in my case).