Resolve double-entry behavior in post editor rooms

[josephfusco]

When a user opens a post editor, two presence entries are written to the same post room:

• editor-{user_id} by the editor heartbeat handler (includes/heartbeat.php:108-116)
• lock-{user_id} by the post-lock bridge (includes/post-lock-bridge.php:49-56)

Both store identical data: same room, same user, same { action: 'editing', screen: ... }. They differ only in the client_id prefix. wp_get_presence() returns both, undocumented, with no way to tell them apart except by parsing that prefix.

The one signal the split encodes is post-lock liveness, since the two rows are refreshed by different heartbeat keys and age out independently.

Proposed fix

Merge into one editor-{user_id} entry with a locked boolean in the state data. The bridge writes into the existing entry instead of creating a row.

Knock-ons:

• Resolves Hidden post-edit tabs still refresh the post-lock presence entry #21. With no separate lock row there is nothing to keep alive in a hidden tab, and PR Fix: Hidden post-edit tabs still refresh the post-lock presence entry #78 closes.
• Removes one write per tick per editing user, changing the baseline Produce a scale analysis and benchmark before the core proposal #132 should benchmark.
• includes/heartbeat.php:80-83 registers lock-{user_id} for the pagehide DELETE. That goes away.

If rejected

#21 is a live bug, PR #78 is its fix and should be unblocked. This issue then narrows to adding an explicit type field so callers stop parsing client_id.

Sequencing

#135 moves this same inline script to a file. Land it after this and PR #78.

[josephfusco]

This issue was identified as part of a peer review conducted by Claude Sonnet 4.6, facilitated by @josephfusco.

[josephfusco]

Refinement note, 2026-08-08.

The Knock-ons section states that this resolves #21 and closes PR #78. That is no longer accurate. #78 was merged on 2026-07-27 with an independent fix, having the ping script drop wp-refresh-post-lock from the heartbeat payload while the document is hidden (assets/js/presence-ping.js:24), and #21 was closed today. There is e2e coverage in tests/e2e/presence-visibility.test.js, the heartbeat-send omits wp-refresh-post-lock while editor tab is hidden case.

So this issue should be judged on the data-model argument alone: one entry with a locked boolean versus two entries distinguished only by a client_id prefix. The hidden-tab bug is no longer a reason to do it.

Worth noting this is on the critical path. #131 states that resolving this removes one write per tick before its own options are evaluated, and #132 needs to benchmark the resulting baseline rather than the current one. Both are waiting on this decision.