To echo Merlin's comment, I would add further the following (or at least my position..).
1. Currently the most convincing argument in favour of deploying DoT/DoH on home routers is to prevent eavesdropping/tampering DNS queries by your ISP. I still remember the guy (
@lancethepants) who mentioned this to me a few years ago when asked about why people want to (or need to) use it.
When an ISP starts deploying its own DoT/DoH servers, it becomes a very interesting/embarrassing dilemma. That is if your last mile is clean, you don't need DoT/DoH. If your last mile is hostile, one greater evil is the ISP...
2. DNSSEC seems to be engineered by people with a greater vision. It'll be good to have and more widely deployed. However, it's not cheap in computation. From my brief tests, it adds noticeable delay. Hence, a better caching DNS server is needed on home routers.
3. DoT/DoH. The overhead between these two at protocol level is trivial and negligible. Performance difference or users' perception of it so far is highly dependent on client/server implementation. I think Dnscrypt-proxy V1 (the one I briefly tried years ago) was horrible. Hence, people perhaps see a noticeable improvement when Stubby showed up. Intrinsically DoH could be just as fast.
The comparison that developers should try to establish is between DoT/DoH and no DoT/DoH. There should two baselines (for comparison) too: one to your ISP's DNS server. another to now IMHO one of those over hyped public DNS servers.
I'm usually pretty good at this sort of analytical stuff. I would expect to see a non-trivial differences. I'm happy to be proved wrong though.