As for the latency - is it really noticeable?
Unbound, recommended for use with Pi-hole, can be configured to log this by enabling "log-replies" in unbound.conf⁽¹⁾ where the time to resolve will be logged in seconds.
⁽¹⁾ https://docs.pi-hole.net/guides/dns/unbound/ ⁽²⁾ https://unbound.docs.nlnetlabs.nl/en/latest/manpages/unbound...
So atleast there's that.
On the go, over 5G, those 12ms won't make much of a difference.
Considering that people deploy PiHole on Raspberry Pi W models, over wifi, you won't lose much running NextDNS, but you gain dns blacklisting on all networks, as opposed to just your home network (or via VPN)
This is my latency (ping.nextdns.io):
zepto-cph (IPv6) 12 ms (anycast1, ultralow2)
zepto-cph 13 ms (anycast1, ultralow2)
■ anexia-cph 13 ms (anycast2, ultralow1) anexia-cph (IPv6) 15 ms (anycast2, ultralow1)so if your dns is slow, there is a tremendous amount of latency added to virtually everything that you do. just because you can hit nextdns in 12ms does not mean the e2e duration for a single dns-then-fetch is going to be in the realm of 12ms. if nextdns doesn't have the answer it needs to go find it.
Yes, if we were in the "good old days" of slim websites, 12ms may be noticable, but today, with webpages taking up lots and lots of storage that is served with every connetion, i seriously doubt you'll notice.
Besides that, every browser and modern operating system will cache DNS records for whatever the TTL from the upstream DNS is set to.