Home › Forums › WoodMart support forum › WoodMart mobile first-image render delay reproducible normal URL, desktop fast

WoodMart mobile first-image render delay reproducible normal URL, desktop fast

Viewing 1 post (of 1 total)
  • Author
    Posts
  • #731953

    tazemgeldi
    Participant

    Hello XTemos Support,

    Please investigate mobile initial rendering on https://tazemgeldi.com/. This is a live WooCommerce store. Preserve its design and shopping functionality. Please perform disruptive isolation tests on a protected staging copy, not production. Do not disable checkout/store plugins or change the active production theme.

    Reported stack: WoodMart 8.6.4 + child theme; Elementor 4.3.6; WordPress 7.1.4; WooCommerce 11.2.2; PHP 8.5.11; LiteSpeed Cache 7.9.3. Redis Object Cache 3.0.1 is connected through AccelerateWP. AccelerateWP page cache is not enabled; LiteSpeed remains the page cache.

    Normal product URL:
    https://tazemgeldi.com/product/metro-chef-surup-icinde-lici-meyvesi-konservesi-2840-g/
    Recent normal-URL PSI report:
    https://pagespeed.web.dev/analysis/https-tazemgeldi-com-product-metro-chef-surup-icinde-lici-meyvesi-konservesi-2840-g/znhw76h0d5?form_factor=mobile
    Mobile: 67/100, FCP 3.0 s, LCP 6.9 s, TBT 30 ms. The exact number varies across runs, but mobile LCP usually remains approximately 6–7.5 s. Another normal-URL run scored 69, LCP 7.5 s, while desktop scored 99 with LCP 0.5 s:
    https://pagespeed.web.dev/analysis/https-tazemgeldi-com-product-metro-chef-surup-icinde-lici-meyvesi-konservesi-2840-g/ircbe2hyqf?form_factor=mobile

    Homepage earlier stable report: mobile 69, LCP 6.1 s; desktop 98, LCP 1.1 s:
    https://pagespeed.web.dev/analysis/https-tazemgeldi-com/zqqjgoycxl?form_factor=mobile
    These are lab results. PSI has no field data for this site.

    Evidence:
    – Mobile LCP is the main product-gallery image; on home it is the first bestseller product image.
    – These images are in initial HTML, eager, fetchpriority=high. The product image is already downloaded well before it renders.
    – In the normal product report the observed LCP breakdown is approximately 49 ms TTFB, 171 ms resource-discovery delay, 266 ms resource transfer, and 1870 ms render delay. These observed values are different from the simulated mobile LCP; they must not be added to the simulated 7.5 seconds.
    – Cached public HTML can respond in roughly 50–400 ms locally. Cold PHP requests still take approximately 2–3 seconds. OPcache 128 MB is full; the host says it cannot raise the limit on this shared plan. Redis does not replace OPcache. Please separate this server issue from the cached-page rendering issue.

    Controlled trials and limitations:
    – Earlier carousel/gallery startup at DOMContentLoaded did not consistently improve normal URLs.
    – Request-scoped JS deferral OFF sometimes gave 95–98 on a diagnostic query URL, but the normal product URL still had LCP around 7.5 seconds. The trial was rolled back.
    – The theme emits wd-scrollbar differently according to mobile/desktop request detection. Forcing the installed theme’s own head script also failed to reproduce a reliable gain on the normal URL.
    – Removing initial carousel transforms did not improve LCP.
    – Removing unused core block CSS reduced uncompressed combined CSS from 424,214 to 283,497 bytes, but LCP stayed about 6.8 s. This was not adopted globally.
    – CSS combination OFF gave 71, LCP 7.6 s; main-image decoding=sync gave 71, LCP 6.9 s. Neither solved the delay.
    – Inlining the same full CSS produced one 77/LCP3.5 result and a later 98/LCP2.3 result. However the unchanged control query ALSO scored 98/LCP2.3, while the normal URL still scored 67/LCP6.9. This does not establish an inline-CSS benefit, so no rollout was justified.
    – Normal and query responses had matching stylesheet/script assets in inspected samples, aside from diagnostic changes and WooCommerce’s random related-product data. Mobile/WebP cache variants must be considered. We have not established a root cause.
    – Temporary probes and timing endpoints are removed before handoff. Please use the archived PSI links as evidence; do not assume old test query URLs still change settings.

    Stable settings:
    CSS minify + combine + external/inline combination ON; async CSS and UCSS OFF. JS minify ON, combine OFF, Deferred (not delayed until interaction). jQuery excluded; gtm.js and analytics.js also excluded from defer. WoodMart CSS-inline limit 40 KB, font-display swap, preloader OFF, experimental mobile DOM optimization OFF. QUIC.cloud is connected; CCSS queue was visible but asynchronous CSS was not enabled.

    Customizations retained:
    Our store helper handles unit/case buying and cart behavior. A separate small helper provides Turkish font glyph subsets with identical shapes/advances. On logged-out home only, it removes Elementor runtime scripts after verifying an allowlist of WoodMart widgets and absence of motion/background effects; unknown widgets retain the scripts. It does not trim product-page Elementor scripts. We can compare without this helper on staging.

    Please advise:
    1. Is there a known mobile first-image/gallery rendering issue in these versions, or a supported patch?
    2. Why could identical assets on normal versus diagnostic query URLs produce such different first paint/LCP? Please check mobile/WebP variants and gallery initialization.
    3. What is the supported way to show the first gallery/card image without waiting for unrelated assets, preserving gallery navigation and zoom?
    4. Which exact LiteSpeed + Elementor + WoodMart settings do you recommend, and are the stated PHP/WordPress/WooCommerce versions supported together?
    5. Can you provide a trace-based diagnosis or identify the smallest staging isolation test needed?

    Please preserve: mobile navigation, automatic brand carousel, two-image gallery, product loading while scrolling, unit/case selection, add-to-cart, shipping and checkout. Do not request credentials in a public forum. Any access request will be reviewed separately.

    Thank you.

Viewing 1 post (of 1 total)