00-moshpit.conf assumed systemd-resolved "uses the first server for every
query, rotating only on failure", so listing DigitalOcean's resolvers
after the bridge was harmless insurance.
It rotates away from a failing server and does not rotate back. One
restart of moshpit-dns moves every query to the fallback permanently, and
the fallback answers NXDOMAIN for every Moshpit name — so DNS looks
healthy while the whole namespace is dead. That is the worst shape a
failure can take: nothing is obviously broken to look at.
Seen three times in one afternoon before it was understood. `dig
@127.0.0.1 -p 5354 seo.rank` answered, `curl seo.rank` did not, and
resolvectl showed the bridge listed and 67.207.67.3 current.
The fallback also buys nothing. The bridge forwards what it cannot answer
upstream itself (mode=clearnet), so a second entry adds no capability —
only that failure. If the bridge stops, restart it: the unit already does,
in two seconds. Two seconds of no DNS beats an afternoon of DNS that lies.
Documents that this must be the only drop-in setting DNS=, and how to
remove the others.