> define "store ordering". Does it affect loads in any way? Or simply just stores

It affects the visible ordering of remote stores to normal memory, so load are necessarily affected (it wouldn't make sense to guarantee a store order if unobservable).

Really, TSO is defined independently of x86 and in fact it took a while to actually prove that x86 was TSO. Concretely, how do architectures that claim (optional) TSO differ from each other at least for access to normal memory?