Etcd is having issues bootstrapping because the bootstrap node by default picks an ipv4 address, which means traffic from the masters comes in from their ipv4 address, which doesn't match what etcd is expecting. See the following error message from the bootstrap node:

{{"rejected connection","remote-addr":"192.168.111.20:48372","server-name":"","ip-addresses":["fd2e:6f44:5dd8:c956::14"],"dns-names":["localhost",
"fd2e:6f44:5dd8:c956::14"],"error":"tls: \"192.168.111.20\" does not match any of DNSNames [\"localhost\" \"fd2e:6f44:5dd8:c956::14\"]"
}}

We don't currently run any node ip selection logic on the bootstrap because it can only have one interface, but since the default kubelet selection logic isn't what we want here we may need to find a way to enable that.