Persistent Geo-location Mismatch: 'What is My Country?' tool misidentifies user IP location on IPv6 networks

hey, we’ve been running our web tool, 'What is My Country? - Find Your Current Country & IP Location', for a while now, and generally it works great. but we're hitting a wall with some really tricky geo-location accuracy issues, especially for specific user segments.

the main problem is a persistent geo-location mismatch where the tool incorrectly identifies the user's country or IP location. this isn't a widespread issue, but it's consistent for users on certain networks, particularly those using IPv6 or mobile data with carrier-grade NAT (CGNAT). it seems our current methods just aren't cutting it for these edge cases, leading to a poor user experience for a noticeable chunk of our audience.

we've tried a few things to combat this. initially, we upgraded our primary geo-IP database, then integrated a secondary one as a fallback. we also implemented client-side JavaScript to get a more accurate time zone and language, hoping it would help cross-reference. for server-side, we're already checking X-Forwarded-For and other common proxy headers, and doing reverse DNS lookups where possible. we even experimented with a few different public IP lookup APIs, but the discrepencies remain.

here’s a typical scenario we're seeing in our internal logs when a user reports an incorrect location:

// Simplified server-side log snippet for 'What is My Country?'
Request-ID: abc123def456
Client IP (from connection): 2001:0db8:85a3:0000:0000:8a2e:0370:7334
X-Forwarded-For Header: Not Present
Geo-IP DB 1 Lookup (MaxMind): Country Code: US, City: Ashburn
Geo-IP DB 2 Lookup (IPinfo.io): Country Code: AU, City: Sydney
Client-Side JS Reported Timezone: Australia/Sydney

Resolved Country (Our Logic): United States
Expected Country (User Report): Australia

it’s clear there's a conflict, and our server-side resolution logic, despite attempts to be smart, is failing to pick the correct one when databases disagree on IPv6 addresses.

so, i'm really looking for some advanced strategies or best practices for improving geo-location accuracy, specifically for IPv6 addresses or scenarios involving complex network configurations like CGNAT or VPNs that might obscure the true origin. are there specific header combinations we should prioritize? or maybe a more robust way to reconcile conflicting geo-location data points from various IP address lookup services?

anyone faced this before and found a reliable solution?