Technical guides to diagnose latency, browser resources, cache policies, CDNs, PHP, queries and performance regressions.
Separate network, browser, origin and third-party delays with repeatable measurements.
Measure DNS, connection, TLS, first-byte wait and download time to locate the slow segment.
Distinguish isolated spikes, cold cache, congestion and saturation with correlated samples.
Find HTTP, hostname and trailing-slash hops that add network round trips before the page is served.
Attribute waits to each external domain before removing or deferring an integration.
JavaScript, CSS, fonts, compression and connection reuse.
Find duplicated dependencies and code that is not needed for the initial view.
Move non-critical scripts off the rendering path while preserving explicit dependencies.
Separate critical styles from styles for pages or components that are not yet visible.
Diagnose FOIT, duplicate font files and delays from external font providers.
Check gzip/Brotli negotiation, MIME types, proxy and Vary headers.
Find requests split across domains and verify connection reuse.
Safe policies, cache keys, invalidation and version consistency.
Avoid recomputing every public visit or storing a response containing another account's data.
Use long-lived caching only for hashed filenames and keep HTML revalidatable.
Reduce unnecessary variants without mixing content that genuinely depends on the query.
Plan selective invalidation, warm-up and traffic control after a purge.
OPcache, PHP-FPM queues, SQL queries and object caching.
Distinguish disabled OPcache, exhausted memory and stale cached code.
Read busy workers and queue depth before increasing concurrency or memory.
Use traces and EXPLAIN with authorised access; do not change indexes blindly.
Review hit rate, keys, TTL and consistency before enabling a distributed cache.
Browser navigation cache and release performance budgets.
Find listeners, unload handlers and policies that prevent restoring a page from memory.
Track transfer size, requests and latency per template with agreed thresholds and repeatable data.