I would be wary of declaring an opening array safe on the basis of the way it interacts with a specific opposing array. The FIDE army happens to have a piece on b/g that starts defended, is worth about the same as a Frog, and can flee without moving a pawn first.
Suppose we shuffled the pieces on both sides with some algorithm similar to FRC. Would that be likely to produce a problematic case like your b/g test, or a calm one like your a/h test?
Put another way, are the problems you had with the Frogs on b/g a special case that is unlikely to recur with different armies, or is every future CwDA army going to need to carefully check to make sure the Fairies can't easily win a piece against them in the opening?