[BUG]

## 🚨 **Please Note** 🚨

To ensure efficient investigation of the issue, please fill out the fields below with as much detail as possible. **Reports that do not follow this template may be closed without notification.** We appreciate your cooperation.

## Describe the bug

The WebSocket connection delivers chunks very slowly until it fails sending at all. Depending on where in the process the connection failed, you can refresh the page most of the time and the GET request will retrieve the full conversation correctly. I have already increased the RAM settings on the websocket handler lambda, but it does not seem too improve the situation.

## To Reproduce

Steps to reproduce the behavior:

1. Run any larger conversation
2. wait for the failure to happen
3. Refresh the page
4. See that the convo is complete, even though the connection broke before

Start by reproducing the failure with a larger conversation and observe the WebSocket handler lambda while chunks slow or stop. Compare the broken connection with the refreshed page's GET request, which reportedly retrieves the complete conversation. Done means large conversations stream reliably over WebSocket and remain consistent after refresh.

Written by the indexing model from the issue text.