Currently, librenms rewrites the values defined in health every time a discovery occurs. For certain devices, health values are set manually and should not be changed, but with each discovery (when a device goes down and then recovers, a discovery is performed again), it overwrites them. For example, we set a high temperature alert, but discovery overwrites them with a lower value, generating false alerts.
It would be interesting to have the option to lock these values, preventing discovery from overwriting them, except in cases where the sensor no longer exists (e.g., a switch that has been replaced with a different model).
We should only overwrite them if the device tells us to and we support detecting them from the device. Do you still want to lock them in that scenario as that seems sub-optimal.