Question 1: Is it correct that in therion and in survex the declination is specified exactly the opposite way?
Yes, it is like that. You are right, there should be explicit warning in the manual.
Therion handbook page 18: true bearing = measured bearing + declination
Survex 1.2.1 manual page 19: (true bearing)=((bearing)-(declination zero err))
So in case that magnetic north is 2° east of geographic north, I have to correct this in therion by -2 and in survex by +2.
I am not sure with this. If you measure 0, true bearing would be 2 (assuming east is 90°), therefore I think you should use 2 in therion and -2 in survex. I assume, your situation is the same, as on this picture http://en.wikipedia.org/wiki/Magnetic_declination.
Question 2: When I enter the measurement data in therion, the meridian convergence is handled automatically (in case a "cs" is specified). However due to historic reasons most of our data is still in survex format. I import the ".3d" file and specify the "cs" during import. I assume that therion is not applying any corrections for the meridian convergence and I need to correct the bearings manually in survex?
It is probably like that. If you want to have precise coordinates in your .3d file according to your CS, all bearing should be corrected for meridian convergence. But I am not sure, if you should add or substract it from magnetic declination :) S.