and the other node shows its handling the sessions?
Is the FQDN that Outlook clients the same as OWA? In other words, since OWA is working, Outlook should be using that same pool I would assume.
If so, then you should take a fiddler trace on one of the Outlook clients when that node is down and see what its trying to connect to.
Another test is attempting to bypass the F5 and set the hosts file on an Outlook client of the load balanced FQDN to the other servers IP rather then the F5, then disable the one that was rebuilt in the F5 and see if the client can connect.
Does it fail for all Outlook clients when that one server is down?
The network team should be able to tell if the clients are using the remaining node once the other goes down as the sessions should increase.