One of the drawbacks of NAT is that it might disrupt services that require the initiation of TCP connections from the outside network or disrupt stateless protocols, such as those using UDP. Why is NAT disruptive in these two cases?

Inbound $($outside to inside connections$)$ depend on the presence of NAT mapping to be forwarded to the receiving host inside the network. NAT mappings can time out and, therefore, prevent the device from translating the incoming packets, which end up discarded.

TCP connections require three$-$way handshake to be completed. NAT allows initial TCP segment, which has flags SYN and ACK set to pass only from the inside to the outside and not from the outside to the inside.

NAT does not translate packets for protocols that do not establish connections and do not maintain the state of the connections.

For applications that use UDP, NAT always translates both IPv$4$ address and port number. Since UDP applications can use whatever port number, it is impossible to make mappings for all; therefore, NAT translations will not happen and UDP segments get discarded.