The “why does Windows not always pick the primary DNS server” “fiasco” is coincidently being discussed in this other thread: DNS and small environments

…this “feature” has been the way it is for a long time.

…my take on it, if I’m not being obvious enough, is you should NEVER specify non-AD-integrated-DNS-servers on end-user devices - because of the chance they will pick another one (2nd, 3rd, 4th, etc.) and then not able to resolve any internal FQDN’s from that point on.

…and your internal DNS server(s) should ALWAYS be “given” to the client via DHCP or via Group Policy - setting the DNS server IPs statically can cause you problems if the end-user device visits other networks (especially wireless hotspots or public networks) and this can also be considered a security risk.

As an aside, mixing different DNS providers (8.8.8.8 vs. 4.4.4.2 vs. your ISP vs. something else) under the forwarders tab on your AD-integrated-DNS-server, I don’t like that either - sure, everyone thinks the more you add makes it “more fault tolerant” which I suppose is true, but your also introducing problems with geo-location and global server load balancing. For example 8.8.8.8 will usually be the best bet to accurately route your YouTube traffic to the nearest Google datacenter, making videos stream the fastest, but your ISP, or 4.4.4.2 (who is probably NOT your ISP BTW) won’t necessary make the best choice, and you can be directed to a geographically further datacenter, and thus adversely affect load times. I would test your upstream DNS server using GRC's | DNS Nameserver Performance Benchmark   or Google Code Archive - Long-term storage for Google Code Project Hosting. and stick with whatever provider is fastest.

1 Spice up