Clouds striking barriers which are not horizontally perpendicular to the direction of travel poses another interesting question. If the cloud is assumed to collide inelastically, and unable to cross the barrier as above, then the cloud will be deflected along the barrier and travel along the barrier at speed equal to 10cos(theta) ft/r, or 1.67cos(theta) ft/s, where theta is the smaller of the angles between the cloud's direction of travel, and the barrier, in radians. If your angle is in degrees, multiply by Pi/180 to get the angle in radians.
Cloudkills that run into long curving barriers are messy, since evaluating the cloudkill's velocity will now require calculus, which I'm not paid enough to get into.
Alternatively, if all of this math is too bothersome, make an arbitrary ruling by fiat. It's not as if the behavior of magical clouds is a well-documented phenomenon anyway.