Finally got a simulation that I like going on. Initial tests with emitting smoke directly was utter fail - there was no way the smoke was going to go nicely hit the wall, sink, then rise up. A liquid sim was done first to provide a base for the smoke fluid to emit from. To get it to rise up, a simple emission of the heat into the temperature field ensured that it would hit the top.
Learnt a new setting on the Iso Offset SOP to convert a SDF volume (the density field from the liquid sim) into a smoke volume with the "Volume Sample" mode. So basically, SDF in, and nice clean smoke volume out. Win!