Fractal flames are created by performing many iterations, the number of which is based on the quality, zoom and dimensions of the image. The first input point used to start the iteration process is random, then all subsequent iterations use the output of the last iteration as their input. This sequence does not continue uninterrupted for the duration of the rendering process, it is instead broken into sub batches. For example, if an image required 1M iterations, it would do 100 sub batches of 10,000.
Each sub batch resets the point to a random location, then begins the iteration process again. This is known as the iteration depth and has the effect of resetting the point trajectory, which prevents it from getting suck on a bad path.
The number of iterations done before resetting the point is the sub batch size. Most fractal flame renderers use a default value of about 10K and do not allow for adjusting it. Fractorium allows changing the value via the Iteration table on the Flame tab, but in practice it results in little change. The only reason you might want to adjust it is if your image does not appear converged enough.
Either way, you can see the feedback immediately so if it appears changing the value is having no effect, then leave it at the default of 10K. Whatever value you use is saved in the flame Xml.