Vertical Player — QA Summary (Rounds 2 & 3)

1. Video #4 Freeze (iSport Desktop) Fixed

The video-stall probe played 6 consecutive videos on iSport homepage (desktop Chromium). All 6 videos played ~5 seconds each without any stall events. Video at index 4 specifically: stalled = false, played 5.0s.

On iPhone WebKit, video index 4 did stall — but this is attributable to WebKit emulation limitations (see Finding 5).

2. Keyboard Navigation (iSport Desktop + e15) Requires Prior Click

The keyboard-nav probe tested arrow-key navigation in two phases:

On iSport, keyboard nav did not work even after clicking the player area. On e15, it required a click first — classic focus-gating behavior. The probe summary confirms: "Quirk confirmed."

3. Ad Entry Offset — Video #2 (Desktop + iPhone) No Issue

The ad-entry-offset probe checked whether ads appear at video slot #2 (the expected first-ad position) or are skipped.

Both runs confirmed ads appearing at the correct entry slot. Ad selectors matched: iframe[src*="doubleclick.net"], iframe[src*="googlesyndication.com"], iframe[id*="google_ads"].

4. Playlist Loop (iSport Desktop) Inconclusive

The playlist-loop probe attempted 12 advances through the iSport carousel (12 cards). The playlist did not loop back to the first video.

Navigation stalled at video index 5 — the probe visited indices [0, 1, 2, 3, 4, 5, 5, 5, 5], getting stuck after index 5 and never reaching videos 6–11 or looping back.

5. iPhone Ads — WebKit Emulation Inconclusive

The iPhone WebKit run on iSport detected ad network activity:

• 262 ad events captured (131 requests, 131 responses)
• Ad-entry-offset: ads appeared correctly at video #2
• Ad-pipeline probe: reported "No ad requests detected after advancing past first video" (status: fail)

The ad-pipeline failure is contradicted by the ad-entry-offset finding ads at index 2 and the 262 captured ad events. This suggests the ad-pipeline probe's detection heuristic may need calibration for WebKit.

Additionally, 3 of 6 videos stalled on iPhone WebKit (indices 2, 4, 6 — all even-indexed videos showed 0s playback). This is a WebKit emulation artifact.

6. Cross-Site Summary Overview

Reflex: All 5 failures stem from the VP carousel not being found on the Reflex homepage. The discovery probe reported: "Carousel not found with selector .video-card--video-vertical. Article player: false." The tag-url probe also failed (function not available after 20s polling). This indicates Reflex either does not deploy VP on their homepage or uses a different integration path. Only the performance probe passed (FCP=1196ms, TTFB=287ms — fast).

e15: Best-performing site. All functional probes passed. Playlist loop confirmed (8 unique videos, loops at advance 9). Two "inconclusive" results are video-init (paused state on load — expected for autoplay policies) and navigation (generic nav probe — superseded by keyboard-nav which passed).

7. Category Page — iSport /fotbal No Player

The discovery probe on isport.blesk.cz/fotbal confirmed: no VP carousel present on category pages.

All 13 probes returned not_applicable (skipped due to no player found). Carousel selector .video-card--video-vertical matched 0 elements. Article player: false.

8. iPhone early playback / autoplay freeze Likely Confirmed

Simona reported: "videos freeze about 1 second in, need a screen tap to resume." The iphone-early-stall probe tested 4 consecutive videos under iPhone WebKit emulation.

Result: initialPlaybackStarted: false for all 4 videos. The stall monitor never triggered because currentTime never progressed past 0 within the 8s wait window. The probe technically reported status: pass (zero stalls counted) but the underlying data shows playback never actually started.

9. iPhone ad consistency (per-video presence) Cannot Reproduce Today

Simona reported the same day (2026-04-09) that on her iPhone 12 mini only the first video showed ads — subsequent videos played without ads. The ad-per-video-presence probe classified ad appearance across 5 consecutive videos.

Result: pattern: 'all', videosChecked: 5, adsObservedCount: 5. Every video 1–5 matched ad selectors (iframe[id*="google_ads"], gpt- containers). Our WebKit run did not reproduce Simona's "ads only on first video" observation.

10. Ad vs. content audio state Inconclusive

Simona reported ads playing silently while content played audibly. The ad-audio-state probe attempted to capture audio state during an ad phase and a subsequent content-only phase.

Result: status: inconclusive. Ad phase captured at video index 2: muted: false, volume: 0.6. The content-only phase was never reached within the 3-video budget — because ads appeared on every video (see Finding 9), the probe never caught a pure content-playback sample for comparison.

For Marian — Recommended real-device checks Action Items

The Round 3 findings above rely on Playwright WebKit emulation, which does not replicate iOS power management or real Safari's media pipeline. To validate, please run the following on your iPhone 12 mini:

1. Low-power mode OFF: open iSport homepage — do videos start playing automatically, or do you need to tap the screen first? (This is the new "early stall" question from Finding 8.)
2. Low-power mode ON: same question — does auto-play behave differently when low-power mode is enabled?
3. Ad audio check: when an ad appears, is it muted or audible? Compare against the content video's audio immediately after the ad ends. (Finding 10.)
4. Screen capture: if you can reproduce "only the first video has ads," please record a 30-second screen capture of that session. We need the timing data to debug the ad pipeline. (Finding 9.)