[change] Change to config check mechanism #643
Description
This is not for new contributors and is intended as a placeholder for future developments.
The config check has limited efficacy: it only detects whether the device config hasn't been applied, but it doesn't detect whether the config agent is not running anymore (crashed, stopped).
The lack of data in the timeseries DB is also forcing us to remove the checksum requests from the recovery detection mechanism in the work we are doing in #641.
It would be great to find an efficient way to store a point in influxdb everytime a device requests a checksum.
This is somewhat also related to #271, as a last seen property should take into account both checksum requests and monitoring requests.
Although at that point we could discuss if it makes sense to continue separating these concerns and whether we could have one agent which can handle different concerns with 1 HTTP or websocket request (eg: hearthbeat, config checksum and monitoring at the same time).