after a bit of testing (6 months coff coff)
I noticed that the problem seemed to be the fact that when compiling a big package the computer gets really hot.
lm_sensors report 85 - 90 C, laptop dell xps m1210 dualcore 2GHz.
After buying a cooling base, the max temp (during those long compilations) went down to 75C.
No more unresponsive system. wii!
Maybe because of the high temp, the system would get lots of failed computations and the overhead of the automatic correction was hogging the CPU. (or maybe I'm just imagining things).
Actually I remembered this solution, because I have a PIII 800, and I remember one while emerging KDE (gentoo around ~2003 building from a stage 3) I would get these kernel panics completely freezing the system.
Since I was working on the shell with no real time feedback of the system's temperature, I didn't realized what the problem was.
Then (because of a strange noise) I remembered to open the box and noticed that the CPU fan was almost stopping.
Apparently needed some deep cleaning. Both the CPU and PSU fans were full of dust to a stopping condition.
After a bit of cleaning a a tiny bit of oil they were flying smoothly again.
And that's how the story goes.
I'm writing this because when I have problems I try to find solutions to similar problems.
Let's not keep it to ourselves, as it might be useful to someone else.