同步系统和异步系统容错率的思考

经常听到有文章引用下面这句话：在同步系统中容错率可以达到50%，而在异步系统中容错率下降到33%。
实际上这句话是有问题的，早在1985年Michael J. Fischer, Nancy Lynch, 和Mike Paterson已经证明，在一个完全异步的系统中，没有任何一种共识算法可以容忍哪怕一个错误节点，史称FLP不可能性证明（FLP是他们三个人名字的缩写）。所以，上面那句话应该把“异步系统”修正为“部分同步系统”。
那么问题来了，到底什么是同步系统，异步系统，和部分同步系统呢？

同步系统（Synchronous System）

援引《Introduction to Reliable and Secure Distributed Programming》第46页中的权威定义：

1. Synchronous computation. There is a known upper bound on processing delays. That is, the time taken by any process to execute a step is always less than this bound. Remember that a step gathers the delivery of a message (possibly nil) sent by some other process, a local computation (possibly involving interaction among several layers of same process), and the sending of a message of some other process (possibly omitted).
2. Synchronous communication. There is known uppder bound on message transmission delays. That is, the time period between the instant at which a