I may not be properly understanding the intent of the change, hence my request for clarification. I haven't fully researched the related jiras. I'll quickly explain host based tokens so you can easily respond.
As you probably already know, the default token implementation always resolves the host to an ip for the token's service. This is fine if the user explicitly specified an ip (maybe due to dns errors, or no dns entries), but the host based tokens preserve the exact hostname or ip the user specified. Preserving the hostname shields the user against ip changes which is very important, for example, when a local or remote NN's ip may change during an upgrade.
That said, will this change be used exclusively by the client to select an outgoing network interface? That will be fine. My concern is if the client starts always using ips for a remote host since the token will probably have to contain the ip – depending how/where these new methods are used. Resolving an ip back to a hostname is insufficient since the user may have provided a CNAME and a reverse lookup will return the A name. If the CNAME's ip changes, the client will not detect it which defeats the host based tokens.
Thanks in advance for taking the time to explain.