Roberson said there are redundancies in the system to avoid these sorts of problems, but the BearMail server and the Web server are a part of the same cluster, and the two machines coincidentally experienced separate hardware failures on the same day, creating the problem.