Incorrect vserver being chosen

Hi all. Suppose I have the following IP Forwarding vservers in my LTM (v17.1.0.3):

ltm virtual vserv-coa {
    creation-time 2023-10-30:08:44:12
    destination 10.0.0.0:1700
    ip-forward
    ip-protocol udp
    last-modified-time 2024-01-25:07:54:02
    mask 255.0.0.0
    profiles {
        fastL4 { }
    }
    serverssl-use-sni disabled
    source 10.10.40.0/24
    source-address-translation {
        pool coa-snat-pool
        type snat
    }
    translate-address disabled
    translate-port disabled
    vlans {
        internal
    }
    vlans-enabled
    vs-index 2
}

ltm virtual vserv-outbound {
    creation-time 2023-10-11:15:24:29
    destination 0.0.0.0:any
    ip-forward
    last-modified-time 2023-10-11:15:24:29
    mask any
    profiles {
        fastL4 { }
    }
    serverssl-use-sni disabled
    source 10.10.40.0/24
    translate-address disabled
    translate-port disabled
    vlans {
        internal
    }
    vlans-enabled
    vs-index 4
}

Why is it that when my nodes 10.10.40.5 and .6 send RADIUS CoA traffic using udp/1700 that the LTM chooses to use the less specific vserv-outbound vserver rather than the more specific vserv-coa vserver?

Based on K14800, the defined order of precedence for NEW connections should be matched based on most specific destination, most specific source, most specific port. In this case a sample connection would be:

destination: 10.100.1.234

source: 10.10.40.5

port: udp/1700

vserv-coa: Destination = 10.0.0.0/8, Source = 10.10.40.0/24, Port = udp/1700

vserv-outbound: Destination = *, Source = 10.10.40.0/24, Port = *

Both my SE and F5 TAC are saying that because there are existing UDP connections matching the vserv-outbound vserver then all other UDP traffic will just end up using that vsever as well. This makes absolutely zero sense. All CoA flows are 100% unique connections, so why would the LTM not evaluate them individually? Please, someone help me make it make sense.