Same issue here. Surely vCenter shouldn't be sending out two different DNS requests from the same ephemeral port before the first request has been replied to.
My assumption is that it is closing the port on vCenter once it gets the first response (from which ever of the two responds quickest), so that the second response hits a closed port and then vCentre sends the ICMP port unreachable.
Packet Capture and related log of an example instance from me:
3168: 13:41:37.191045 802.1Q vlan#111 P0 10.216.20.28.36228 > 10.200.223.6.53: udp 58 3169: 13:41:37.191152 802.1Q vlan#111 P0 10.216.20.28.36228 > 10.200.224.6.53: udp 58 3170: 13:41:37.192739 802.1Q vlan#111 P0 10.200.224.6.53 > 10.216.20.28.36228: udp 128 3171: 13:41:37.193639 802.1Q vlan#111 P0 10.200.223.6.53 > 10.216.20.28.36228: udp 128
Jun 13 2020 13:41:37 %ASA-4-313005: No matching connection for ICMP error message: icmp src oob-lzb:10.216.20.28 dst oob-met:10.200.223.6 (type 3, code 3) on oob-lzb interface. Original IP payload: udp src 10.200.223.6/53 dst 10.216.20.28/36228.
Looks like a bug to me.
We are running vCentre Server Appliance 22.214.171.124000 (with embedded PSC)