Telegram plugin (0.0.1) crashes on network interruption

Description

The Telegram channel plugin (claude-channel-telegram@0.0.1) crashes silently on any network interruption (VPN reconnect, Wi-Fi switch, brief connectivity loss). Once crashed, the MCP server disconnects and cannot be restored without restarting Claude Code.

Root Cause

In server.ts (line 594), bot.start() is called fire-and-forget with no error handling:

void bot.start({
  onStart: info => {
    botUsername = info.username
    process.stderr.write(`telegram channel: polling as @${info.username}\n`)
  },
})

Missing:

1. No bot.catch() middleware — grammY throws unhandled errors on polling failures (ETIMEOUT, ECONNRESET), which crash the process
2. No reconnection logic — once the polling loop dies, the bot never reconnects
3. No process.on('uncaughtException'/'unhandledRejection') handlers — any unhandled error kills the MCP server process

Steps to Reproduce

1. Start Claude Code with the Telegram plugin active
2. Disconnect/reconnect VPN or switch Wi-Fi networks
3. Telegram MCP server disconnects within seconds
4. Tools (reply, react, edit_message) become unavailable
5. Only fix is restarting Claude Code

Expected Behavior

The plugin should survive transient network issues and automatically reconnect polling to the Telegram Bot API.

Suggested Fix

bot.catch((err) => {
  process.stderr.write(`telegram channel: bot error: ${err.message}\n`)
})

void bot.start({
  onStart: info => {
    botUsername = info.username
    process.stderr.write(`telegram channel: polling as @${info.username}\n`)
  },
}).catch(err => {
  process.stderr.write(`telegram channel: polling died, restarting: ${err.message}\n`)
  // reconnect logic here
})

Environment

• Claude Code v2.1.80
• Plugin: claude-channel-telegram@0.0.1
• OS: macOS (darwin/arm64)
• Network: corporate VPN (NetBird) with periodic reconnects