It is normal that there is no consensus (yet) about the unstable trailing part.
If we denote the interval of time between a transaction issuance and the transaction finalization (by witnesses) as
confirmation time, does an average confirmation time increase monotonically if the global TPS rate increases monotonically?
PS: "Yes" would mean that SatoNatomato was wrong that DAG has no limits on scaling even if we have superpowerful hardware (but the latency still persists).
It all depends on the behavior of witnesses. That said, it is reasonable to expect that as TPS increases, witnesses also post more frequently, therefore confirmation time
decreases. Of course, there is a lower bound defined by network latency, times number of witnesses.