Routing is quote different between AFriH and AfriG if you look at the above traces I have posted.
If the port was reset - why did the hop come down to 12ms suddenly and then go back to +500ms or trace 2?
If the exchange was congested I would never get 12ms on the DSLAM.
Routing is exactly the same between the two traceroutes, the 2nd just repeated a hop when you reached our website cluster. We don't run multiple networks, so everything from radius to routing is done on one network for both @afrihost.co.za and @afrigreen accounts.
Congestion can happen for a number of reasons, and a port reset is known to ease congestion temporarily.
Anyway, are you able to replicate the results using another ISP?