This may or may not be fixed with

• fix(agent): Close stdin and stdout separately to fix pty output loss coder#6862

Edit: we are still seeing network stalls occasionally and this mostly presents itself in the form ofdd timing out when uploading the worker.

There is potentially another timeout that somehow only occurs when you connect too quickly after a failure? Might just be a coincidence.