* handle the off part of on with timed off command
* use correct var
* only bind / configure cluster once
* clean up channel configuration
* additional debug logging
* add guard
* prevent multiple discoveries for a device
* cleanup and still configure on rejoin