We have a set of web servers that we believe can comfortably handle simultaneous users. However, we only have enough application instances to serve simultaneous users. We have the option of adding more servers to scale the application, but this will be costly, and we don't see it as being necessary at the moment. However, suppose something unexpected happens, and more than the usual number of users all turn up at the same time: the web servers will cope, but the application servers will struggle, and everyone will experience poor performance.