fix(broker): increase healthcheck start-period, catch Grammy errors

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

apps/broker/Dockerfile

@@ -37,7 +37,7 @@ COPY --from=deps --chown=bun:bun /deploy /app
 
 EXPOSE 7900
 
-HEALTHCHECK --interval=15s --timeout=5s --start-period=10s --retries=3 \
+HEALTHCHECK --interval=10s --timeout=5s --start-period=30s --retries=5 \
   CMD bun -e "fetch('http://localhost:7900/health').then(r=>{process.exit(r.ok?0:1)}).catch(()=>process.exit(1))"
 
 # Non-root user (oven/bun image ships with 'bun' uid 1000)

apps/broker/src/telegram-bridge.ts

@@ -1404,13 +1404,15 @@ export async function bootTelegramBridge(
     pushHandler,
   );
 
-  // Start Grammy long-polling
+  // Start Grammy long-polling (fire-and-forget, must not crash broker)
   console.log("[tg-bridge] starting bot...");
   bot.start({
     onStart: () =>
       console.log(
         `[tg-bridge] bot running — ${meshConnections.size} mesh(es), ${chatMeshes.size} chat(s)`,
       ),
+  }).catch((err: unknown) => {
+    console.error("[tg-bridge] bot.start() error:", err instanceof Error ? err.message : String(err));
   });
 }