forwarding round-robin fails/ends on first server?

[juju4]

I had a setup with 3 internal nameservers to which dnscrypt was forwarding the corresponding domain.
When outage happened and one server was unavailable, dns queries were failing once in three times and back to normal when outage ended.
If during outage, I remove the offending nameserver, dns queries are working fine again all the time.
So, it seems that for dns forwarding

• nameserver order is not relevant
• queries are round-robin between each nameserver
• if failing on the given nameserver, others are not tried (round-robin or not)

is it correct?
is there an option to make it try other nameservers if defined and first one fails?
Tested with dnscrypt 2.1.5 (pending new test on current 2.1.12)

[jedisct1]

The order is important for $FALLBACK and $DHCP. But when there's a set of IP addresses, queries are spread randomly across them.

[juju4]

Thanks @jedisct1 but if query fails on first try, does it try other nameserver(s)? or a configuration option to make it try?
it does not seem so and did not find such an option.