Parent DS checks should probe all configured servers even if some results are negative
Different DNS resolver implementations have different approach regarding TTL of the delegation as well as TTL of the DS record. By actively probing various different implementations (BIND, Unbound, Knot Resolver, Google Public DNS, Quad9, Cloudflare), one can make sure the KSK is not rolled too early.
The current logic in Knot is however done in a way that if first probe returns negative result, the rest of servers is not queried. This leads to a situation where cache of the other servers is kept empty and when the first probe finally succeeds, the others will immediately pull fresh DS records from the authoritative servers, leading to immediate positive probe result.
It would be better to keep probing all configured servers, so cache of all of them is kept filled with the old DS record and all probes will succeed only when all caches got their content refreshed.