I did try looking up those gearslutz post concerning “negative” delay - some of those are indeed quite old. Those threads are claiming that certain tracks are being recorded earlier than they should be. They seem to be confined to specific audio interfaces. Somebody quotes RME as saying “negative delay values are possible, Steinberg is aware…”
There you go - RME driver reporting a “negative” delay and Cubase ignoring it. Both companies have an argument here.
For one, “negative” delay can only be a relative measurement, not an absolute measure since it is difference between two measurements. For a relative measurement you can have either the same, more, or less. Which one is the reference of the two measurements determines if this relative measurement is negative or positive.
Delay itself though, is a absolute sum, you either have some of it, or none at all. No reference is needed, since the absence of any delay is simply 0.
RME is reporting a negative delay, they say. Really though, compared to what? If compared to 0 that would be logically impossible, for obvious reasons. So Cubase appears to have adopted the stance that, since negative system delays are not possible, ignore any driver that reports such a thing.
I would be willing to bet a wooden nickel that the early times on these tracks are precisely that “negative” delay*-1 …