Might help to clarify terminology related to the BIGIP:
A node is what you may typically refer to as a server. In your example, you have one node (10.68.15.28).
A pool member is a service, or a server + port. In your example you have two pool members: 10.68.15.28%10:8081 and 10.68.15.28%10:8091
Note that a node may map to several pool members, and can even do so in the same pool as you are trying to do here.
By default, BIGIP will list service names for ports (as defined in /etc/services). This can be changed with the command:
tmsh modify cli global-settings service number
(Doesn't appear that a service restart is required. Last I looked this up, it was a db variable that required a service restart. Just tested this on 12.1.2.)
As for your original request, having two pool members associated with the same node in a single pool is perfectly fine and doesn't require anything outside of setting them up as normal pool members. When you say "the application will not work on both ports", do you mean you cannot access the application through the virtual server when both pool members are enabled? Can you be more specific in the error/symptom you are receiving and the testing you have conducted?
You mentioned that you "did bypass and it is working", so I assume that your client can successfully connect directly to both pool members (both ports on the server). There could be an app-side issue that requires persistence but I can't say without knowing the app.