That probably means there wasn't a good testing process for patching and there wasn't adequate redundancy. In theory, if a patch breaks one server it shouldn't matter.

In reality, patch testing stacks up and gets behind and redundancies are rarely tested. That is expensive, time consuming work which probably isn't worth the time of someone who is already underpaid and overworked. And fuck! If patch and redundancy testing ever breaks anything prod for whatever reason, the person who was testing everything gets blamed and fired so nobody is going to volunteer for that.