It's not going to matter where your origin is hosted on that changes the path to CF, but the network sending (well, mostly). Although in this case I suspect it's not a routing to CF issue, it looks like they're sending some amount of free plan/non-enterprise requests to other data centers after they hit AMS, probably capacity related
You can play around with it a bit yourself to see, send 10 or so requests to a free plan website (you kind of already have), and see what you hit via /cdn-cgi/trace or CF-Ray response header, and then look at mtr and see if you're always on the same route to the local DC. Then repeat w/ an Enterprise website like cloudflare.com. If they both have similar paths in mtr which don't change, and only some amount of requests are rerouted on free, safe bet it's just capacity based redirecting, nothing you can do about directly, should resolve in time though. I'd assume they closed out the incident since the network level is fine now at least for requests actually handled in ams, and the ones being redirected are just seeing normal latency