Because the plane takes off and lands on the same runway, this mean %50 of the time its pointing
towards the pilot, and %50 of the time its point away and because of that I can neglect direction of travel.
That depends on the 'statistical' likelihood of the direction of failure. If you assume that the model is either pointing towards you or away from you (and my models only truly 'point' at me for an extremely short duration on each circuit) and the model failure direction exhibits a Gaussian distribution (for example) then I am not that likely to get hit at all (thankfully). If this same distribution is assumed when flying away from me then I am even more happy.
My whole point of mentioning this is that you can easily reduce the probability of hitting yourself (or someone near you) by the way that you fly and the way that you position the model. This is the basis of most of the air display rules that are written to prevent you crashing into the crowd. I know you are simplifying the problem to indicate that, in any case, this likelihood is low but it is not showing the full story. This is, again, the limitation of statistics.
There are lies, damn lies, and statistics - Mark Twain.