What is the Telegram multi-account polling avalanche bug?

When running 16+ Telegram bot accounts, all accounts simultaneously stall on getUpdates (~110-136s timeout), then cascade-restart, overwhelming the Gateway with concurrent WebSocket reconnections.

What causes the Telegram polling avalanche?

No polling stagger between accounts, no restart jitter, and Gateway cannot handle 16 concurrent reconnections. All accounts start/timeout/restart at the same time.

How to work around the Telegram polling avalanche?

Reduce the number of concurrent bot accounts, or use the per-account polling stall threshold override to stagger timeouts.

Does the polling avalanche affect single-bot setups?

No, this specifically affects multi-account setups with many (16+) Telegram bot accounts running simultaneously via long polling.

Is there a fix planned for the Telegram polling avalanche?

The fix involves adding polling stagger (random offset per account) and restart jitter so accounts do not all timeout and reconnect simultaneously.

← Back to dashboard

clawsmith.com/signal/openclaw-telegram-multi-account-polling-avalanche-16-bots

⚠ IssueWide OpencoreLive

Telegram Multi-Account Polling Avalanche: 16 Bot Accounts Stall and Cascade-Restart Simultaneously

Running 16 Telegram bot accounts via long polling causes simultaneous getUpdates stalls (~110-136s timeout) with no polling stagger, followed by cascading restart that overwhelms Gateway with 16 concurrent WebSocket reconnections and transport rebuilds.

Product Idea from this Signal

A runtime watchdog that isolates OpenClaw plugins into monitored worker threads and kills any plugin that blocks the gateway event loop for more than 5 seconds

OpenClaw plugins (especially Active Memory, Dreaming, and provider bridges) run in the same Node.js event loop as the gateway. When one plugin blocks, it starves everything: Telegram polling stops receiving messages, API responses go blank, and conversations fail silently. Seven bug reports in April 2026 alone document cascade failures from a single plugin blocking the loop. This watchdog spawns each plugin in its own worker thread with a heartbeat monitor, kills and restarts any plugin that exceeds the event loop latency threshold, and logs the exact stack trace that caused the block.

CLIOPEN-SOURCEDEVTOOLRUNTIMENODE.JS

CompetitiveView Opportunity →