DNS-over-HTTPS was "allegedly" made to secure both privacy and security concerns, f.ex. a MITM attack as UDP sends in plain text (as mentioned above where an ISP can snoop up DNS request and then sends their own DNS response, regardless).
However, as usual, it can provide more problems than it solves:
You can read more about it here:
https://www.sans.org/white-papers/39160/
Main points are:
DoH doesn't actually prevent ISPs user tracking
DoH creates havoc in the enterprise sector
DoH weakens cyber-security
DoH helps criminals
DoH shouldn't be recommended to dissidents
DoH centralizes DNS traffic at a few DoH resolvers
https://blog.mozilla.org/en/products/firefox/firefox-continues-push-to-bring-dns-over-https-by-default-for-us-users/
Right. Firefox family tends to use CloudFlares DoH resolvers, but LibreWolf (the one I use) has it off (if I remember correctly). Not sure about Chrome since I don't use it. But yeah, it's easy to forget you've made changes to either that or the hosts file which may screw the results later on (as Timppu found out).
timppu: 1. How does CDN decide to which server you will be connected?
2. Why does it seem different CDN servers appear to be configured so differently?
CDN's relies heavily on back-ends and load-balancing, but as I mentioned before, what route it takes through the different networks depends heavily on available routes and configurations.
Traffic to/from DNS servers and traffic to/from servers with the other protocols (i.e. file transfers) isn't the same, but yes, depending on what you get from the DNS it can be pretty bad, especially if the network or the server you finally get is configured badly, has high traffic, or limits its bandwidth.
With Wireshark and Nmap you'll be able to find a lot of information about the path through the different networked tiers, however, load-balancing and backends are generally something that happens "behind the scenes" while only exposing one or a handful of IP addresses.
If Wireshark or Nmap is too daunting, then I can recommend at least
tracert (Traceroute), a good tool to in every network troubleshooting:
Here's what I get when I try
tracert 151.101.85.55 4 19 ms 19 ms 20 ms irb-2925.agg2.xxx.us.m247.ro [37.120.220.38]
5 25 ms 25 ms 26 ms te-0-0-0-5.bb1n.xxx.se.m247.ro [37.120.220.158]
6 * * * Request timed out.
7 26 ms 25 ms 26 ms 151.101.85.55
And with
gog-cdn.us-eu.map.fastly.net I get:
4 19 ms 21 ms 19 ms irb-.-.us.m247.ro [37.120.220.38]
5 18 ms 18 ms 20 ms ae6-.-.ip4.gtt.net [212.222.111.25]
6 26 ms 26 ms 26 ms ae3.-.ip4.gtt.net [89.149.129.98]
7 * * * Request timed out.
8 27 ms 27 ms 27 ms be4649.-.-.atlas.cogentco.com [130.117.3.130]
9 35 ms 33 ms 33 ms be4094.-.-.atlas.cogentco.com [154.54.37.58]
10 35 ms 34 ms 35 ms be4090.-.-.atlas.cogentco.com [154.25.16.154]
11 32 ms 32 ms 32 ms 149.6.117.26
12 32 ms 32 ms 34 ms 151.101.237.55
You see how the traffic is diverted depending on if you use the host name or the IP directly? The same goes for different protocols and apps and the order of things as well (nslookup f.ex. completely bypasses the hosts file while ping -a doesn't).
So, subdomains and IP addresses of those sub-domains can change any time depending on where you are, what DNS you use and get from it/them, and the configuration of that network. And using VPN won't necessarily be better.
This really is down the rabbit- (PI-)hole. I need coffee...
EDIT: I know where .ro is supposed to be, but all I see is
.ro
.ro
.row your boat... XD