Docked→handheld transition leaves the surviving instance bound to the external pad — Deck built-in controls locked out

[aradanmn]

Found during PR #78 (2b) on-Deck validation, 2026-07-10 (operator: aradanmn)

The docked→handheld transition itself works as designed (validated live: DISPLAY_MODE_CHANGE → screen re-probe → slot 2+ teardown → slot 1 survives and reflows fullscreen, isolation intact). But the surviving instance keeps its docked-mode controller binding — the external pad (DS4) that claimed slot 1 at docked launch. The Deck's built-in controls are structurally excluded from player slots in docked mode (by design, so they never steal a slot), and nothing rebinds them on the transition.

Net effect for a real user: you undock to move to the couch and the handheld only responds to the external pad you presumably just walked away from. The built-in controls are locked out until you quit and relaunch in handheld mode.

Direction

On the docked→handheld transition, the surviving slot should transition its input to the Deck's built-in controls (which is what a fresh handheld launch binds). Options, roughly in order of plausibility:

• Restart-slot-1-on-transition with handheld binding semantics (matches the restart-on-reconnect direction in Instance sandboxes leak other players' input nodes and go stale on re-enumeration — isolation is app-layer only (probable #61 mechanism) #62 — rebinding a live bwrap sandbox isn't possible, same constraint).
• Or treat the transition as teardown-all + fresh handheld spawn (simpler, but P1 loses their session — worse for a LAN host).
• The Seamless controller reconnect via a per-slot stable virtual device (input proxy) #38 input-proxy design would make this trivial (re-attach the built-in's events to slot 1's stable virtual device) — consider making this a named acceptance case for Seamless controller reconnect via a per-slot stable virtual device (input proxy) #38.

Milestone suggestion: v1.2 (rides the #62/#38 input-binding work).

Refs: PR #78 validation session; the transition path is orchestrator.sh DISPLAY_MODE_CHANGE→handheld.

[aradanmn]

Evidence from the 2026-07-15 PR #82 validation session: after a docked 2P session was carried through an undock (P2 reaped, session continued), the built-in Deck controls were dead in all instances — the surviving instance was still bound to its external DS4's nodes, exactly this issue's mechanism. A fresh handheld launch afterward worked perfectly (built-in via its Steam virtual, direct-launch path). Also relevant: during the same session's docked disconnect/re-plug test, slot reclaim worked at the STATE level (same slot reactivated, no new-slot grab) — the gap is purely the input path (#62 static binds), which strengthens the case that the #38 per-slot virtual device is the right v1.2 fix for both this and reconnect.