I've noticed that if I build a cluster rocket, but use different motors in the different positions, (for example a triangular cluster with 2xB6-4 and 1xC6-5, the rocket still flies straight "up" (when there is no wind). It looks like the thrust is being calculated as the sum of the three engines, but all acting at the centerline of the rocket (axissmyetrically). In the real world, there the thrust vector would be off-axis so there would be an imparted picthing moment.
The real reason to simluate this behavior is for the case where the engines are symmetrical and equal thrust, but they ignite at slightly different times - or in the even of a large cluster rocket where one or more engines fails to light at all.
This is a known issue that I'm planning on addressing later once I've refactored the motor handling in general. Currently the simulation has no way of separating different motors of a cluster, that is desireable in various cases.