It is important to choose an IP address range that does not have or has goods odds against, conflicting with remote client address pools. If VPN uses IP addresses from a range shared by a remote client address pool, packets meant for the client LAN may attempt to traverse the VPN to the wrong system or to a system that doesn't exist at all. Alternatively, the traffic may never leave the client LAN and be routed to a local resource, instead.
The following diagram illustrates a fairly severe case of what I'm describing. There are various resources identified with their associated LAN address on both sides.
On the left, there is a network where the VPN server resides. The LAN on the server network uses the 10.4.0.0/24
subnet. For the VPN, the 10.8.0.0/24
subnet is used. This will facilitate VPN traffic, and a route will be pushed for the server-side LAN subnet. There are two internal servers for which the VPN was created. The first is an application server using LAN IP 10.4.0.76
...