I think this probably needs to be in another issue, but I can't locate an open one, and I don't want to open a new one until I know that there's not an existing one we can do the work in:
The startup logging, when in cloud mode, should log the precise information that will be used when registering with zookeeper – host, port, context path, http/https, etc. Have a user on IRC that has run into a problem starting a second node because it sees an existing ephemeral live node that matches the info it's trying to use to register. Setting the "-h hostname" option on bin/solr fixes the issue. If the logging had included the precise registration info, it would have been easier to track down. Ultimately we figured out that both hosts were trying to register as 127.0.1.1 – because /etc/hosts was incorrect.