I model DoT effects as if they are continuous rather than coming in discrete ticks:

Modeled DoT potency = [Full DoT potency] × MIN([active length of DoT] ÷ [Full DoT duration], 1).

The core assumption here is that the ticks have an equal chance to be synced to any time (and since they start pulsing outside reasonable player control, this is sane enough, I think). Thus, the portion clipped is exactly how much DoT potency you should expect to lose over an infinite number of runs. So, losing the last tick because you clipped by 0.01s is treated like the freak accident it is.