receiving, by a processing device executing a wide area network (WAN) accessible service, a firmware update for a plurality of devices, wherein the firmware update is to be provided to the plurality of devices by an over the air (OTA) update;

identifying, using a selection criteria, a set of candidate devices from the plurality of devices to receive the OTA update, wherein the selection criteria comprises one or more device attributes;

determining a status check time interval based on historical usage patterns associated with the set of candidate devices;

determining, by the processing device executing the WAN accessible service, one or more idle state conditions that indicate an idle device state for a class of devices associated with the set of candidate devices, wherein the one or more idle state conditions comprises the status check time interval;

selecting a candidate device from the set of candidate devices to receive the firmware update; and

providing, to the selected candidate device, the one or more idle state conditions and the firmware update, to cause the candidate device to monitor, using a machine learning classifier, for the idle state conditions before applying the firmware update, wherein the one or more idle state conditions are associated with an expiration time period associated with applying the firmware update regardless of the idle state conditions of the selected candidate device.