As the source IP of connections originated in the UTM, the IP of the outgoing interface to the destination is used.
IPSec has no own "interface", so the interface bound to the IPSec connection is used (usually the WAN interface).
So, if the destination network is reached via an IPSec connection, make a SNAT
WAN (address) -> Destination LAN --> LAN(address)
Connections originating in the UTM are AD authentication, SUM connection, syslog, proxy, mail notification ...
As the source IP of connections originated in the UTM, the IP of the outgoing interface to the destination is used.
IPSec has no own "interface", so the interface bound to the IPSec connection is used (usually the WAN interface).
So, if the destination network is reached via an IPSec connection, make a SNAT
WAN (address) -> Destination LAN --> LAN(address)
Connections originating in the UTM are AD authentication, SUM connection, syslog, proxy, mail notification ...
Not all ipsec connections go through a WAN connection, we are running it through a dark-fiber in a private vlan. What we are seeing is the connection of the remote UTM originating from 1 of the defined VLAN's, not sure if it is random selected, last one created or based on internal identifier, but it tends to change when a VLAN is added (and maybe on reboots).
So SNAT would be a nice option, but just like the firewall rules we currently use, the SNAT would also break on the change of the source address. This is why it would be nice to have the function that creates the route for the ipsec would use the route metric of the vlan interfaces to determine which one to use....