Event bus worker does not retry on startup failure

## Summary If the event bus worker fails to start (e.g., database not yet reachable via Toxiproxy during container startup), it logs the error and never retries. The event bus remains permanently non-functional until the container is restarted. ## Reproduction Using the resilience test bed (`tests/resilience/`): 1. Start the Docker stack — occasionally, mesh starts before Toxiproxy's Postgres proxy is fully ready 2. Mesh logs show: `[EventBus] Error during startup: connect ECONNREFUSED 172.19.0.6:5433` 3. After this, events are published to the DB but never delivered (worker is dead) 4. Only a container restart fixes it ## Expected Behavior The event bus startup should retry with exponential backoff (e.g., 1s, 2s, 4s, up to 30s) until it succeeds. This is critical for: - Container orchestration where service ordering isn't guaranteed - Rolling deployments where the database connection pool may be temporarily saturated - Transient network issues during startup ## Impact - **Silent data loss**: Events are published and stored in Postgres but never delivered. No further error logs after the initial failure — the system appears healthy. - **Production risk**: In Kubernetes, pod startup order is non-deterministic. If the DB connection pool is briefly saturated during a rolling deployment, the event bus dies permanently on affected pods. ## Suggested Fix In `apps/mesh/src/api/app.ts` (line ~788), wrap the `eventBus.start()` call in a retry loop: ```typescript async function startEventBusWithRetry(eventBus, maxRetries = 10, baseDelayMs = 1000) { for (let attempt = 0; attempt < maxRetries; attempt++) { try { await eventBus.start(); return; } catch (error) { const delay = Math.min(baseDelayMs * 2 ** attempt, 30_000); console.error(`[EventBus] Startup failed (attempt ${attempt + 1}/${maxRetries}), retrying in ${delay}ms:`, error.message); await new Promise(r => setTimeout(r, delay)); } } console.error("[EventBus] Failed to start after all retries"); } ``` ## Found By Resilience test bed: `tests/resilience/docker-compose.yml` — observed during initial container startup when Toxiproxy wasn't ready before mesh