As discussed in [1], we should be making the "file" store more robust for it to resist to common failure scenarios due to the location being remotely mounted via NFS

For example:

• handle i/o timeouts
• ensure the remote filesystem is available before writing/reading data

1. https://docs.google.com/document/d/14QQ3hyMRAAP9ZRIpbwsyklxRhjq0MsGF9Mj8dRCw3dY