I'm an evidence-inclined person. Earlier today I wanted to check propagation times on the BCH mainnet, so I summoned a fleet of 256 nodes all around the globe and registered the times at which each transaction was first seen. So I collected the times of arrival of each transaction at each node and then compared them. The test was ran in a span of some 2 hours since 19:30 CET of Sunday 31/05/2020
The result is in the title: https://i.imgur.com/dsfDCGb.jpg. We've always known this, but it's cool to have reproducible evidence.
I anticipate that this methodology can be used to test various other aspects of the distributed network, like propagation times for blocks, performance of graphene/xthinner et al., behaviour under stress/high volume, pre-consensus, etc.