I would fix this issue with the "MAT/TPS enrichment" table.
With a closed or near closed throttle, increase the enrichment until you reach lambda target.
I've solved it this way on a couple of engines, usually BMW M30 due to the intake pipe going across the valve cover to the hot exhaust side.
There is usually no need to increase enrichment in the table for higher throttle openings, as the sensor does not heat soak in that case.
Thanks for your answer.
I've already tried to play with this correction table without success. You're right in that the problem is for idle only, not at light or heavy load.
Based on my previous log I should add 10% more fuel for 36C than 56C.
-sounds extreme for density correction, and this direction is just the opposite of correcting heat soak, right?
But if I look into another log made a few minutes before the first one,
http://quasar.dynaweb.hu/~lezsi/vems/v3.3_n002211-2010.07.31-15.59.42.vemslog-I see, that from 30C to 44C MAT with 2.5ms inj pw results the same lambda (0.94) at idle.
So, for this just-warmed-up situation adding 10% more fuel won't be a solution.
That's why I said that I can't spot any real differences in measured values to base on.