Reboot loop during machine configuration application in Talos Linux

Talos Linux utilizes a gRPC API for cluster management, replacing traditional shell access with YAML-based machine configurations applied via talosctl. During the bootstrapping phase, the apply-config command pushes the desired state to the node to transition it from maintenance mode to a running state.

A critical failure point occurs when invalid kernel arguments or incorrect disk partition definitions are included in the configuration. Because the OS lacks a local shell for manual intervention, a misconfiguration that prevents the network stack or the API from initializing can result in a continuous reboot loop.

Given the API-driven nature of the platform, how can a recovery path be established when the network configuration is broken? What are the documented mechanisms for overriding a corrupted machine configuration when the gRPC endpoint is unreachable?