The first probe is synchronous so the very first readiness answer is evidence rather than the pessimistic default — otherwise every deploy reports one interval of false negatives and orchestrators learn to ignore the probe.
Fixed delay and not fixed rate: when the broker is unreachable each probe takes the full admin timeout, and a fixed rate would queue overlapping runs against a dependency that is already struggling.