OK the problem seems to be elsewhere, without actually seeing cpu load or network load, the syslog server running on the NUC was being hit by a vm sending thousands and thousands of logs... seems stable now. Odd that I didn't see any other side effect than CPU frequency increase.
For those running a NUC have you witnessed some cpu frequency related issues?
To avoid having the thing "run" at full pace, I'm using the powersave governor, and recently updated to 5.3.13, 5.3.18 and 5.4.24 but all of these kernel do not seem to respect the instruction (pstate issue?).