Interestingly they used to run on IIS, which is why you saw a lot of ebayisapi.dll in their urls.
Can't imagine ebay having any problem moving to new URLs nor getting any significant boost in referrals from such actions. What other backwards compatibility is at issue, scraping apps?
I am surprised EbayISAPI.dll was C++ - I always assumed it was a mess of .NET. It makes sense considering how old Ebay is though.
I see over and over again the 'no database would scale big enough, so we had to build our own'. If only opensource databases got spanner-style auto-sharding and auto-loadbalancing sooner, millions of engineer-hours could have been better spent!