IP address lookup accuracy

We operate a 'What is my IP Address' web tool, and our primary focus is on delivering highly accurate IP address lookup results. Recently, we've been observing notable discrepancies in the reported IP addresses and their associated geolocation data, particularly for users operating behind intricate network configurations. This issue becomes pronounced when dealing with nested proxies or various CDN setups, where the true client IP can be obscured or misrepresented.

Our core technical hurdle lies in reliably parsing and prioritizing HTTP headers such as X-Forwarded-For, CF-Connecting-IP, and X-Real-IP when multiple intermediaries are in play. The challenge isn't just about identifying *a* public IP, but specifically the *true* originating client IP, distinguishing it from an intermediary's address, especially in scenarios involving long proxy chains or IPv6 tunnels which further complicate robust client IP detection. We're looking for insights into what constitutes the most robust, battle-tested server-side logic or perhaps a specific library for accurately extracting the real client IP address from a request, taking into account varying header permutations and the ever-present threat of spoofing. How do others manage the hierarchy and validation of these critical headers to ensure a truly reliable IP address lookup?