Tumble dry : low
Over the weekend, I have added the Reaction Control Sytem (RCS) of the re-entry module (SA) of the Soyuz, based on the few data I could gather from either pictures or documents. And, as you could have guessed, there's much room left for guessing. The SA is only supposed to be able to orient it-self before (and during?) re-entry, so it is equipped with 6 thursters (9.85N each). These allow the Soyuz commander (middle seat) to adjust the pitch, roll and yaw during the brief time from separation of the service module (PAO) and parachute deployment (about 12 minutes).
The position and thurst vector of an attitude control thruster is something that is studied with lot of attention during the design phase of any spacecraft. The idea been that, the linear moment created by a thruster need to be cancelled out by another thurster located at the opposite position. In the case of the SA, this rule doesn't seems to have been applied, with only one set thrusters (2 each) by attitude components (pitch,roll and yaw):
While testing the SA in LEO, I have noticed that the roll thrusters are generating some unwanted moments (ehence the slow tumbling). I said unwanted ... but in fact I don't know if it is... What is happening make sense to me, but as we have based our implementation of the re-entry module on sketchy data (thrusters location was guessed from pictures as well as their thrust vectors), it's hard to know if the end result is an accurate simulation ... I guess we need to dig around for more data ... :-