I think what Invictus means is run the proxy against the thing you actually care about and watch the numbers, not a score from some site. Send a few hundred requests to your real target (or something close to it) and track success rate, how many captchas/403s/429s you get, and median plus p95 latency. A proxy can look clean on a fraud score site and still get blocked on your target, and the reverse happens too.
For stuff you can check up front, I'd do a leak test (no X-Forwarded-For or Via header added, and DNS/WebRTC not showing your real IP if you're in a browser), then look the IP up in two or three different geoip/ASN databases. If they disagree on country or city, or one calls it hosting and another says residential, better to know before you use it for anything geo sensitive. Also test it a few times over a day. With rotating pools quality varies a lot between IPs, so one sample tells you almost nothing.