GeoIP Precision Discrepancy Woes

Building on previous discussions about Geolocation API issues, my SaaS heavily depends on precise IP-based location data for compliance and content delivery, and we're currently facing a critical technical block with inconsistent data. We're seeing significant, inconsistent discrepancies in GeoIP precision across multiple commercial providers like MaxMind and IPinfo.io, and even within our own internal database. Approximately 15-20% of our users are being mislocated to adjacent cities or even entirely different states, which is frankly unacceptable for our use case that requires regional or even sub-city level accuracy. We've certainly attempted several solutions: we've cross-referenced data from several leading GeoIP databases to try and find common ground, and we've implemented client-side browser geolocation as a fallback, but it's largely unreliable due to user blocking and simply not suitable for robust server-side validation. We've also conducted network route analysis, using tools like traceroute and MTR for problematic IPs, observing patterns with large ISPs routing traffic through central hubs, but this hasn't yielded any clear, actionable insights for correction. Exploring custom IP range analysis for known problematic regions was another avenue, but this approach is clearly not scalable for our user base. We even investigated VPN/proxy detection, but the core discrepancies persist even for seemingly 'clean' IP addresses. Lastly, we looked into DNS-based geo-load balancing, but our fundamental issue is with the IP-to-location mapping itself, not just routing optimization. The specific technical challenge we're grappling with is the need for a robust method to programmatically reconcile conflicting location data from multiple reputable GeoIP providers. Are there established statistical approaches or heuristics to weight providers based on their historical IP geolocation accuracy in specific geographic regions or network types? How can we effectively identify and 'correct' common misattributions that plague large ISP IP blocks? We are desperately seeking advanced strategies, architectural patterns, or specific algorithmic recommendations to significantly enhance IP geolocation accuracy for critical server-side operations, particularly when faced with ambiguous or conflicting location data. Has anyone successfully tackled this specific GeoIP precision challenge at scale, and what were your most effective solutions?