fix(realtime): terminate web worker on disconnect to prevent memory leak

[tanmaysharma2001]

🔍 Description

This PR fixes a memory leak in RealtimeClient when using the worker: true option. Web Workers were being created during connection but never properly cleaned up during disconnection, leading to memory accumulation over time.

What changed?

1. Added _terminateWorker() private method

   • Terminates the Web Worker when called
   • Clears the workerRef reference for garbage collection
   • Includes logging for debugging

2. Updated _teardownConnection() method

   • Now calls _terminateWorker() during connection teardown
   • Ensures workers are properly cleaned up on disconnect

3. Added test coverage

   • New test verifies workers are terminated on disconnect
   • Ensures workerRef is cleared after termination

Why was this change needed?

When using RealtimeClient with worker: true, each disconnect/reconnect cycle would create a new Web Worker without terminating the previous one. This caused:

• Memory accumulation with orphaned workers
• Workers continuing to send heartbeat messages even when disconnected
• Potential performance degradation over time with multiple reconnections

The fix ensures proper worker lifecycle management by terminating workers during connection teardown.

Closes #1902

📸 Screenshots/Examples

Before: Workers were created but never terminated

// _teardownConnection() - line 628-649 (before fix)
private _teardownConnection(): void {
  // ... cleanup code
  this._clearAllTimers()
  this.channels.forEach((channel) => channel.teardown())
  // Worker never terminated - memory leak!
}

**After**: Workers are properly cleaned up
```typescript
// _teardownConnection() - line 628-650 (after fix)
private _teardownConnection(): void {
  // ... cleanup code
  this._clearAllTimers()
  this._terminateWorker()  // Worker properly terminated
  this.channels.forEach((channel) => channel.teardown())
}

🔄 Breaking changes

• [☑️] This PR contains no breaking changes

📋 Checklist

• [☑️] I have read the Contributing Guidelines
• [☑️] My PR title follows the conventional commit format: <type>(<scope>): <description>
• [☑️] I have run npx nx format to ensure consistent code formatting
• [☑️] I have added tests for new functionality (if applicable)
• [☑️] I have updated documentation (if applicable)

📝 Additional notes

[coveralls]

coverage: 95.367% (+14.2%) from 81.183%
when pulling 6f6fec5 on tanmaysharma2001:fix/realtime-worker-memory-leak
into 8d1e0c5 on supabase:master.

[tanmaysharma2001]

Hi @filipecabaco, I wanted to ask regarding the Label Issues and PRs check. I am not completely sure why is it failing. Is it something i need to fix from my side? Please let me know. Thank you 🙏

[mandarini]

@tanmaysharma2001 don't worry about this. I think sometimes it fails when you do not accept edits from maintainers, but don't worry about it. maybe i will remove this workflow in the future, if it fails consistently.

[mandarini]

Great fix! This correctly addresses the memory leak by terminating workers on disconnect.

One related issue: the error handler in _startWorkerHeartbeat() has the same problem. If a worker errors out, workerRef isn't cleared, so on reconnect no new worker gets created. Could you update the error handler to use your new method?

  this.workerRef.onerror = (error) => {
    this.log('worker', 'worker error', (error as ErrorEvent).message)
    this._terminateWorker()
  }

[tanmaysharma2001]

Hi @mandarini, i have updated the said changes. Can you have a look please?