I don’t know anymore what this could be. The version p11 is ok the stale problem was until p10.
And only with active checks 
- the data is sent from the agent
- If you do a “cmk hostname” on the command line the service is refreshed
Only point left
- is there a rule for “Normal check interval for service checks” active who is affecting the ntp?
Look at the “Parameters for this service” and there at the “Monitoring Configuration” section.
Compare this section with the “Parameters for this service” of the “Check_MK” service.
Is this all the same?