Seems to be one legged configuration, i.e. virtual server and pool members (nodes) are in the same subnet.
Might be a routing problem for the return packets from the servers. If you monitor the packets to the server do you see any traffic going to the servers and anything coming back? You can check with GUI: Statistics ›› Module Statistics : Local Traffic ›› Virtual Servers or Nodes: Bits/Packets In /Out. If you see something going in but 0 out, then the return packets may bypass the BIG-IP and connection cannot be made.
Do the servers have a route back to the client via the BIG-IP? Where is the servers' default gateway pointing? If not to the BIG-IP self IP (1.0.0.1), do you have any SNAT, like SNAT Automap for the virtual defined? SNAT Automap will rewrite client address to be the BIG-IP's address for the server bound packets. Thus the servers think they are communicating with the BIG-IP on packet level, and will send the return packets back to the BIG-IP, which will pass them back to the client.