I always recommend reverting to the OSI model to structure troubleshooting attempts.
Start with layer 1 but from how the issue is described, perhaps there's a layer 3/4 component? hosts.deny, hosts.allow, and iptables ties to layer3/4 troubleshooting. However, ssh is an application so there may be application-related issues as others have mentioned.
So after C is refused, B can't connect to A. Can you describe more detail what "B can not ssh to A too" means? Can you ping A from B? Can you ping A from C? What about pinging from A to both B and C?
Are A, B, and C on the same subnet? Are they on different subnets?
Is this behavior tied to specific user accounts? Are you using password auth? Have you tried cert-based auth?
Reviewing the pam config and sshd config would also help. What do you see in the log files?
Did this config ever work? Did you make any changes that could have broken something?
Every detail helps when posting such a question, especially on a forum.