Part 2 of the ERTH Architecture Series: How to build a Bun watchdog daemon to detect, kill, and resurrect crashed Python sidecar processes with dynamic port negotiation.
In the first part of this series, we established the core foundation of our ERTH Stack (ElectroBun + Robyn + Turso + HTMX) desktop application. We successfully spawned a high-performance Python backend (Robyn) as a child process from a Bun master process, using Port 0 to dynamically allocate a port and avoid port collisions.
But in real-world desktop environments, things are rarely that simple.
What happens if the Python process enters an asynchronous deadlock? What if it crashes due to a C-level memory access violation? Or what if the operating system aggressively reclaims memory and terminates your sidecar backend while the frontend shell remains open?
If you do not have a robust recovery strategy, your application will hang indefinitely, leaving your user staring at a frozen, unresponsive screen.
