Our demo server (demo.opennms.org) can stay up and running for around 3 days before start having memory issues, after upgrading to OpenNMS 14.
Checking the logs, there are OutOfMemory exceptions, there are also consecutive Full GC, and there are several auto-generated heap dump in /opt/opennms.
After analyzing all the heap dumps with Eclipse Memory Analyzer, I found that all the leak suspects are related with Vaadin applications, especially the OpsBoard feature, even if it is not configured to show data.
There are some known memory issues related with the same classes reported as leak suspects fixed on Vaadin 7.2 that could fix this problem.
Right now, demo is running a branch of 14 using Vaadin 7.2, and the OpsBoard has been configured to show some data.
If the server stays up for more than 3 days, that means, Vaadin 7.2 fixes the problem.