For approximately 120 seconds, the device gateway was not accepting software manifests from devices when they checked for an update. After a restart of the affected service (initiated automatically), normal processing resumed.
No data was lost, and any devices that checked for an update during that window would have been able to retry a few minutes later.
The total length of the incident was initially reported as 4 minutes, but upon further investigation we were able to determine that the service was down for at most 120 seconds.