He has one process that always takes the left-most branch and one which always takes the rightmost (and they then do standard depth-first search within that) - but I think the others are all just set to random. You could probably weight the searches to tend to the left or the right and then have a spread of such tendencies.
I'm fairly sure they're not doing that but, now you've suggested it, it seems a fairly obvious direction to go in, in order to try and improve coverage.
no subject
I'm fairly sure they're not doing that but, now you've suggested it, it seems a fairly obvious direction to go in, in order to try and improve coverage.