Function health 

pub async fn health(__arg0: State<Arc<AppState>>) -> impl IntoResponse

Handles GET /health — returns readiness status, worker counts, and tuning diagnostics.

Returns 503 while models are loading or if all workers have exited; returns 200 ok (or 200 warn when fewer workers are live than configured) with the current cost-model coefficients and probe status in the tuning block.