Persistent IP geolocation accuracy issues post-MaxMind GeoLite2 update causing data resolution errors

hey folks, we've been running our IP Lookup Tool for a while now, and recently we upgraded our backend database to the latest MaxMind GeoLite2 version. the idea was to improve overall data freshness and accuracy for IP geolocation, especially with all the dynamic IPs out there. however, since the update, we're seeing some pretty weird and persistent accuracy issues.

specifically, for certain IP ranges – think mobile carriers or some regional ISPs – the IP geolocation data is just wildly off. sometimes it's showing a different city or even country than what we were getting before, or what other reputable services report. it's not a complete failure, but it's inconsistent, making our tool less reliable for those specific segments. we've double-checked our integration, the database path, and permissions, everything seems fine on that front. here’s a simplified console output example of what i'm seeing:

# Querying a known mobile carrier IP (e.g., 100.20.30.40)
# Before MaxMind update:
#   City: New York, Region: NY, Country: US, ISP: T-Mobile
# After MaxMind update:
#   City: Ashburn, Region: VA, Country: US, ISP: Unknown
#   Confidence Score: Low (0.6)

# Querying another problematic IP (e.g., 172.50.60.70)
# Before MaxMind update:
#   City: Los Angeles, Region: CA, Country: US, ISP: Verizon
# After MaxMind update:
#   City: Miami, Region: FL, Country: US, ISP: AT&T
#   Confidence Score: Medium (0.75)

it's really frustrating because these discrepancies are impacting our users. i'm looking for insights on advanced debugging strategies for such data resolution problems. has anyone faced similar common pitfalls during MaxMind database migrations or updates? are there specific validation techniques beyond just spot-checking that could help us pinpoint why certain ranges are misbehaving? any thoughts on how to reconcile these conflicting data points without rolling back? anyone faced this before?