In reply to Offwidth:
> (In reply to bpmclimb)
>
> Maybe think harder. You are looking at an unlikely combination of the knot aligned perpendicular to the tension, failure to have redundancy and no stopper knot. Its not that its impossible its just that there are so many other factors more likely to be important. Of course it is irresponsible for some not to waer a helmet when belaying at the top of a crag, or use less than 4 pieces each with locking crabs.
The fig 8 (tied in both strands together rather than rethreaded) is generally considered unsafe for joining two ropes for abseil because it can roll. There is a strong consensus to support this, supported by experimental evidence.
My point (and that of the OP, I think) was that in the limiting case described - only a fig. 8 tie-in loop used for both anchors and belay plate, and no stopper knot (not really such an unusual scenario, as anchors and belay plate are supposed to be opposed in direction, and a stopper generally not regarded as necessary for the integrity of the knot) - the forces on the knot will be similar. Why then is the fig 8 tie-in loop almost universally regarded as safe as an attachment point? Why is there no (apparent) tendency for it to roll? I agree with the OP that this is a question worth asking. (I still think this even after thinking further about it).
Your point about redunancy is, I think, not relevant, because the tie-in loop of a fig 8 is not only regarded as safe, but often actively recommended as a superior choice to the harness loop. Yet if the knot were to roll off the end of the rope, the whole system would fail.
Your points about numbers of anchors, use of locking crabs, and the wearing of helmets are all well and good, but don't really shed any light on the particular question raised by the OP.