There is definitely a break-point for when QoS (or even a
single component of QoS like traffic engineering) makes sense.
Someone asked earlier in this thread if it was cheaper to add
capacity or pay for the bright engineers to make TE or QoS work.
For large carriers, the right answer is often to pay for the
bright engineers.
I did some traffic engineering work for a major IP carrier a
few years ago. The cost savings from traffic engineering at the
time more than paid for the cost of the quality engineers they
had. Of course, the same amount of bandwidth they had then
would now cost much less, and be considered a small network,
and the results today could well be different.
Prabhu