Indeed, sshd service does not close active sessions while shutting down.
With a killall sshd, you would be shutting down your own sshd server as well, which is no big deal if it is already stopped, though may be risky, working from a remote station.
To avoid this, I would first locate PIDs for sshd instances bound to a client:
# who am i root pts/0 2019-10-25 13:52 (1.2.3.4) # ps fxww|grep pts/ 12144 ? Ss 0:00 \_ sshd: root@pts/0 12150 pts/0 Ss 0:00 | \_ -bash 12205 pts/0 R+ 0:00 | \_ ps fxww 12206 pts/0 S+ 0:00 | \_ grep pts/ 12169 ? Ss 0:00 \_ sshd: root@pts/1 12175 pts/1 Ss+ 0:00 \_ -bash
Now I know I can kill 12144 or 12169 closing existing sessions. COnsidering that my session is attached to pts/0, I would probably avoid killing 12144.
To automate this:
exclude=`who am i | awk '{print $2}'` ps axww | grep -v "$exclude" \ | awk '/sshd: [^ ]*@pts/{print $1}' \ | while read pid; do kill $pid done