The docs for the /reserve endpoint say:
This is not true: the master returns 200 when the request has been validated and a CheckpointResourcesMessage has been sent to the agent, but the master does not attempt to verify that the message has been received or that the agent successfully checkpointed. Same behavior applies to /unreserve, /create-volumes, and /destroy-volumes.
We should either:
1. Accurately document what 200 return code means.
2. Change the implementation to wait for the agent's next checkpoint to succeed (and to include the effect of the operation) before returning success to the HTTP client.