ISPs don't allocate IPs sequentially so if you ban a range it can be anywhere in that ISPs region - the wider the range ban the more potential users impacted
The IPs belong to Exponential-E who look to be more a Microsoft house than AWS so chances are the IPs are either Front Door (load balancer) or some firewall (Palo Alto or the likes). As for the number of VMs, or Web App instances or Function apps etc behind it know way to know. They may limit the scaling to keep costs down and would explain some of the issues faced.
I wonder if they could be external IPs of a WAF or other Firewall or something rather than a load balancer as tracert dies when it hits them. Looking up the hosting provider I am going to guess it's cloud based and using scale sets but no easy way to see what's behind the point of entry. Certainly looked round robin as I hit 4 different IPs on the morning all in the 31. range - looking on my hosts file from previous sales days 167. range was definitely used in the past as that's what I had saved.
All very interesting if not definitive π