Still having this issue on OpenFire 3.9.3 and spark 2.6.3 Running on CentOS 6.6. Pulling Roster List from Active Directory on server 2012 r2.
Exiting spark and re opening it will correct for a few hours. Any idea why I should look into and start tweaking? I have been playing around with the cache settings and the seemed to help other issues where my roster would remove people on its own.