A remote processor restarts on an embedded Linux board: what should the host application do?

Treat the restart as a loss of remote state. Pending commands, sequence numbers, shared buffers and configuration may no longer match the host's assumptions. Track a session or generation identity and require a defined initialization handshake before resuming normal traffic.

Separate processor lifecycle handling from application message semantics. The framework can manage firmware and remote resources, but it does not decide whether an interrupted command should be retried. Design idempotent operations or explicit acknowledgement and reconciliation for actions that must not be repeated blindly.

Test resets during each meaningful protocol phase and verify resource cleanup on the host. Bound reconnection attempts and expose a degraded state instead of waiting forever. For outputs that affect hardware, define the safe state during loss of communication and the conditions for restoring operation.