Hey, thanks for the detailed report. The bcId and repro steps really help.
This is a known bug, the same one as in this thread: Cloud agent reports error after initialization without additional information. The agent does bring up the environment and load the snapshot correctly, but the kickoff message gets lost before the conversation loop starts, so StreamConversation keeps failing with Missing request message. This is a server-side issue, not related to your repo or environment, and not related to your network or VPN.
We’ve reported the issue internally, but there’s no exact ETA yet. Once we have an update, we’ll post it in the linked thread.
As a temporary workaround, you can try running the agent from the IDE via the Agents window on the same branch and repo. That uses a different kickoff code path and often works. Some users are still seeing it fail from the IDE too, so it’s not guaranteed. If you try it, let us know if it helped.