when you've a Marathon in a HA setup, you certainly expect one to go down, but what IP address to you connect to then when current one fails and remains down.
solution 1
discover all members via DNS SRV record (host and port), connect to any of them, use it; if one fails, try another one until it succeeds.
solution 2
allow passing multiple endpoints via command line parameter, and connect to one of them; if that fails, rotate through that list until one succeeds.
The endpoint list may contain DNS names to allow runtime relocation through DNS A record changes. (keeping the same port)