if you count timeouts as an improvement

Have you seen this yet? https://github.com/jmcorgan/fips/blob/master/docs/design/fips-intro.md

Interestingly I think IPV4 is better for DHTs because the scarcity makes Sybil attacks more difficult. To have the same effect on an IPV6 DHT, I imagine you need to increase the redundancy factor to maximise the chance of finding an honest node, but that adds cost too.