Hello,

On November 30, 2025, between 13:05-13:19 UTC, my Telegram bot

stopped responding to user messages. Messages were delivered (double

checkmarks in Telegram), but the bot did not process them. I had to

manually restart the service at 13:19 UTC.

This has never happened before.

Railway metrics during the incident:

- Memory: spiked to 3.2 GB before the incident

- Response Time: peaked at 20 seconds

- Request Error Rate: peaked at 6.5%

- After restart, Memory dropped to ~400 MB

PostgreSQL logs:

- Database remained stable throughout the incident (checkpoints

every 5 min)

- At restart time (13:19:38 UTC), 20 DB connections dropped

simultaneously ("Connection reset by peer")

- No deadlocks, lock timeouts, or slow queries in DB logs

Application logs:

- Last activity: 13:05:50 UTC

- After restart: 822 tasks marked as timed-out

- High activity before incident (many concurrent handler calls)

My questions:

1. Was the service killed by Railway due to resource limits (OOM

killer, CPU throttling)?

2. Does Railway have an automatic restart mechanism for unresponsive

services, and did it trigger in my case?

3. What is the memory limit for my plan, and was it exceeded (3.2

GB)?

Configuration:

- Service: Telegram Bot (Python, python-telegram-bot)

- Database: PostgreSQL (separate Railway service)

- Plan PRO

I would appreciate your help in determining the root cause of this

incident.

Thank you.