Telegram error 429: Too Many Requests: retry after N

Your bot exceeded one of Telegram's sending limits and must pause. The description names the wait in seconds and the same value is returned machine-readably in parameters.retry_after, which the Bot API defines as 'in case of exceeding flood control, the number of seconds left to wait before the request can be repeated'. The number varies per response; 429s seen in the wild range from single digits to tens of thousands of seconds for sustained abuse.

The FAQ publishes the guidance this enforces: avoid more than one message per second in a single chat (short bursts may be tolerated), no more than 20 messages per minute in a group, and bulk notifications to no more than about 30 messages per second overall. All three throttles are collected with sources under Telegram rate limits. Paid Broadcasts in @BotFather raise the bulk ceiling to 1000 per second for a per-message fee in Telegram Stars.

Libraries turn this error into a typed exception with the wait attached. python-telegram-bot raises telegram.error.RetryAfter with a retry_after attribute; aiogram v3 raises aiogram.exceptions.TelegramRetryAfter with retry_after; grammY exposes err.parameters.retry_after on the GrammyError; Telegraf exposes err.response.parameters.retry_after; node-telegram-bot-api leaves it in error.response.body.parameters.retry_after. Whatever the stack, the contract is the same: sleep at least that many seconds before retrying that request, and slow the overall pipeline, because retrying early typically extends the penalty.

The request itself was valid; the chat, token and content are fine. Telegram is asking you to slow down, not telling you something is broken.

{"ok":false,"error_code":429,"description":"Too Many Requests: retry after 34","parameters":{"retry_after":34}}

Put every outbound call through one rate-limited queue that honors retry_after globally, not per worker, and that applies per-chat and per-group budgets on top of the global one. Alert on the 429 rate rather than on individual 429s, which are normal at broadcast scale. The exact documented numbers live on the Telegram limits page, and the launch-day failure pattern is cause 5 in Telegram bot not responding.