Steps to reproduce:

1. Set up "StatsAllowedIP" in server version 6.4 configuration file to allow gathering instance statistics.
2. Go to the item "Zabbix stats" with key zabbix[stats,\\{$ADDRESS},\\{$PORT}]  from "Remote Zabbix server health template" (or create similar item).
3. Press "Test" and fill {$ADDRESS} and {$PORT} of the configured server v6.4, press "Get value and test".

Result:

JSON data returned by the server is in format that differs from previous versions (see "remote_stats_6.4.json" and compare it with data returned by previous versions (I've also attached examples of data queried from servers v5.0/6.0/6.2). Result is the same if you'll use server 6.2/6.0/5.0 to query the item from server v6.4. So in 6.4 "Remote Zabbix server health template" and, I guess, "Remote Zabbix proxy health" are broken because of that.

Expected:

All metrics should be placed under "data" object like in previous versions.

Also, I think it would make sense to place cache metrics (vcache, tcache, rcache, wcache) under one object (like "cache") in the same way it is made for "process" metrics (for all versions since 5.0, but users will have to get updated version of "Remote Zabbix server health template"/"Remote Zabbix proxy health" templates where JSONPath preprocessing will be changed accordingly).