One has to consider what is daily use, not just the exceptional cases.
generally it's the exceptional case that it's designed around though
We have a car bridge designed to withstand exceptional load, and a pedestrian bridge designed to withstand exceptional load. The commenter assumes that on average, the car bridge's load is much closer to it's maximum load than the pedestrian bridge's load, and in consequence, the average wear on the car bridge should be higher than on the pedestrian bridge. As such, the pedestrian bridge should have a much longer lifetime, and the commenter assumes that this is the reason old Roman bridges are still standing.
Their question is: is this assumption correct?
You also have to take failure modes into account, and how degraded the bridge can get before it's "unsafe".
It's a little like memory pressure vs cpu vs disk vs network. There are some services that aren't going to ever hit one of them because the others are limiting first. Memcached is never going saturate disk.
If you've got a bridge that's 'stiff' or 'heavy' (like truss/beam/whatever) resonances are unlikely to be an issue. If you've got anything light and flexible or with cables, you need to be thinking _hard_ about resonance, in all the modes, torsional being one that's bitten designers bad in the past.