I think this is really the most important point. Link speeds and such
are not as significant, maximum packet rates probably are.
Where link rates become key is in modeling the worm's dynamics. Because
often a single infected machine could fill an access link, additional
infections behind that same link didn't change the worm's overall scanning
rate; this is unlike other, non-"bandwidth-limited" worms, and changes
Slammer's growth to not be the usual exponential/logistic curve.
See our paper on Slammer for more details:
We're now trying to explore that effect, which requires knowing what those
bottleneck bandwidth rates were for sets of hosts behind a common bottleneck.