this morning I monitored the frequency shift between the FP-cavity frequency (the laser is locked on it) and the RF reference frequency @ 500MHz,
during the warm-up time, for 50kW stored in the cavity (the recording started ~ 9h20).

one observes on fig.1 a simple exponential behavior with 500Hz frequency shift @500MHz (equivalent to ~ 10µm) for 50kW stored. 
the time constant T of the exponential curve is 12.5 minutes and the stable region starts at 5T ~ 1h.

the small drop at the end of the curve ( ~ @10h30) could come from the external temperature which started to drift before (see fig.2)