Cloudcache in active active datacenter causing Outlook to throw an "old OST" error

So I have cutover a client to a new 2022 Citrix deployment in an active:active datacenter deployment.  We are using cloudcache fslogix to store the profiles in storage volumes in each datacenter.  Each datacenter is set to use the local storage volume first, and the far datacenter second.  

Profile types are set to 3.  RW first, RO subsequent.

This works fine for the majority of the users.  As they launch a desktop, or a single published application and work fine. 

But sometimes users get click happy.  Or they launch a desktop from one machine, disconnect it.  Then launch an application from a different machine.

The issue is when users get access to resources in both datacenters.  If I launch Word from datacenter A it locks the profile RW and works correctly.  But if I launch Outlook in datacenter B it sees the profile is locked and sets it up as RO.  But Outlook throws an error saying that the OST file is an old copy.  This only happens when I connect in the 2 separate datacenters at the same time.  If I launch Word from datacenter A and also launch Outlook from a different server in datacenter A, it creates the RO profile, but Outlook is still able to work correctly.

Next week we will be testing this theory by having both datacenters prefer 1 storage volume.  Luckily the datacenters are pretty high speed connected, so it shouldn't be an issue.  But has anyone else seen anything like this?  

Our next step if this doesn't fix it is to look into only enabling cached exchange mode on proper profile mounting.  It is my understanding that RO sessions don't count as a proper mount, so it should run in online mode only.  Am I correct, or incorrect on this.