Desktop should survive renderer reloads with automatic conversation checkpointing and recovery
Summary
Long-running ChatGPT desktop conversations are vulnerable to renderer reloads, blank windows, and UI rehydration failures. While conversation data that has already been submitted is generally preserved server-side, the active desktop session can lose important client-side state.
Several existing issues describe renderer reloads, blank windows, or stalled UI state. This request focuses on a complementary capability: automatically preserving and restoring the user's in-progress desktop session after renderer failures.
Problem
When the renderer reloads because of a crash, update, memory pressure, or similar event, users can lose:
- unsent composer text
- scroll position
- active streamed response
- pending attachments
- expanded conversation context
- project context
- other UI state that must be manually reconstructed
For users who maintain long-lived research or engineering conversations, reconstructing that context is costly even when submitted messages remain available.
Proposed capability
Implement automatic conversation checkpointing and session recovery.
The desktop application should periodically checkpoint enough local state to restore the active session after a renderer reload.
Examples include:
- active conversation
- scroll position
- unsent composer contents
- streamed response state where feasible
- pending attachments
- active project context
- UI layout and panel state
Following a renderer restart, the application should transparently restore the previous working state whenever possible or clearly indicate which elements could not be recovered.
Benefits
- Prevents loss of in-progress work.
- Makes long-running desktop conversations significantly more resilient.
- Reduces interruption from renderer crashes or reloads.
- Improves confidence in using ChatGPT desktop as a persistent research and engineering workspace.
This proposal is intentionally distinct from workspace export or backup requests. Those address long-term ownership and portability. This request addresses live-session resilience and recovery.