fix: reverse playlist persists across video navigation (fixes #3759, #3818, #3883)

[jiangyj545]

(running playlistReverseUpdate() twice with 1.5s delay)

Fix for #3759, #3818, #3883 — Reverse playlist reverts on video navigation

### Problem When "Reverse Playlist" is active and the user navigates to the next video in the playlist, YouTube's SPA navigation triggers `yt-page-data-updated`. The extension calls `playlistReverseUpdate()`, but at this point `ytd_watch.data` may not be populated yet, so the function returns early without applying the reverse. The playlist then renders in its original order.

Solution

Add a delayed retry (1.5s) in playlistReverse(). After the initial playlistReverseUpdate() call, a setTimeout fires once more to catch the case where the first attempt returned early because playlist data wasn't ready.

This is minimal and safe:

• Only retries when playlistReversed is still true
• Does not add an interval or loop — just one additional attempt
• The existing isCurrentlyReversed === shouldBeReversed guard in playlistReverseUpdate() prevents double-reversing if the first call succeeded

Changes

File	Change
playlist.js	Added 1.5s delayed retry of playlistReverseUpdate() in playlistReverse()

Testing

• 66/69 unit tests pass (3 pre-existing failures unrelated to this change)
• JS-only change, no CSS or HTML modifications

[ImprovedTube2]

hi! @suryaprakash0010 #3766

[Hyperspeed1313]

Try the latest release first? #3643 should've already fixed this

[wahajahmed010]

Review: Reverse playlist SPA fix

Potential issues:

1. Stale setTimeout pileup — setTimeout has no cleanup. If the user navigates between videos quickly, multiple 1.5s delays stack up. The ImprovedTube.playlistReversed === true guard helps but stale callbacks from earlier navigations could run when the user is on a different video. Consider storing & clearing the timeout ID.

2. #3643 overlap — Hyperspeed1313 noted #3643 may already fix this. PR references 3 issues — worth clarifying if they persist after #3643.

3. Magic number — 1500ms works empirically but could be a named constant with rationale.

What looks good:

• Minimal single-file change
• Guard prevents double-reverse
• Real SPA race condition addressed