This problem doesn't exist on H24, Meridian2018 or even early snapshots of H25.
With the current state of the develop branch, there are no Newts statistics when collecting data through JMX from the OpenNMS server, but data collection is happening normally and data is being stored on Cassandra/ScyllaBD through Newts without issues (for other JMX stats, SNMP, etc.).
On hawtio, I cannot see any statistics associated with org.opennms.newts.
Without these statistics, there is no visibility of the injection rate, the expensive indexing phase, resource cache, ring buffer, newts disk usage, etc.
In other words, you'll be 100% blind when using Newts, with no way to know if your settings are correct or not, as these statistics are unavailable on Cassandra or ScyllaDB, as they reflect the Newts usage from OpenNMS's perspective.
For this reason, I marked this issue as a bloquer.