Just decided to sign up to inject some extra weather information into an already compelling discussion, and to back up some of weatherdude's input.
I am an amateur, not a professional, and do not hold any kind of pilot/aviation credentials.
The METAR report is a good source for surface conditions and observations from there, and I'd like to add that the CB report is quite likely correct. Any cumulus cloud (Cu) that rains is by definition a cumulonimbus (Cb), the discriminator for a thunderstorm is in METAR code as TS, with +/- for each to indicate rain rate. The reason for this is that any Cu is technically convective and thus can cause upset, with TCu also used to warn of non-raining but potentially strong updrafts. The METAR reports around the accident time correctly state showers and broken low clouds, hence CB's, with stratiform cloud cover at 10k ft. Given the immediately pre-frontal nature of that night's weather at Rostov this is a likely cloud pattern.
A better source for weather conditions, if you can find one nearby to an accident time/location, is a sounding; it just so happens a balloon was launched from Rostov-on-don at 00z (likely just prior, so perhaps 1hr before accident). The lowest levels up to 700mb are available here:
34731 URRR Rostov-Na-Donu Sounding in which on either side of the diagram you can see the height of a recorded measurement and wind information (bearing, strength) from this you can infer a wind shear difference of 40kts within ~600m of the surface. Additionally, there is slight vertical directional shear (difference in wind bearing with height), both of which would suggest shear-induced wave action, ie: turbulence. This amount of low level shear is quite extreme, and on an airfield already known for turbulence.
The synoptic scenario on the night was a strong frontal zone advancing from the NW that had just reached the Black Sea area at 00z, and was at it's strongest point for that location. Here is the 850mb wind chart for that part of Asia taken from the initial chart of the GFS model (note: this would be mostly assimilated actual data, just gridded, before the model started to compute anything):
http://www.tropicaltidbits.com/analy...850_asia_1.png A quick look on Google maps at the local terrain shows that the wind fetch was coming straight off the Black Sea up into a valley where Rostov is with an elevation change of 78m. Adding to the vertical turbulence effects would have been the possibility of further horizontal eddying from orographic wind effects of the valley coupled with the approaching front maxima. All in all a very bad weather situation to head into.
As for icing, the complete diagram for the 00z Rostov sounding shows saturation between 0C and -20C:
34731 URRR Rostov-Na-Donu Sounding (note for non-weather folk, that diagram is a skew-T which means temperature on the x-axis relates to the blue lines that are the most skewed to the right from immediately above their labels, the black lines are the instrument measurements so to read off the temp see where they intersect at your chosen height) which means riming was possible. It also shows that at above 5.5km there was less saturation (separation between the lines of temp and dewpoint measurements), so quite possible there was minimal icing risk in the hold (esp being off to the SW of this reading) but perhaps some icing in the descent to the 2nd attempt.
brb an associate is running a GFS-WRF simulation at 2km grid scale