I'm currently maintaining an application written by a former co-worker. The entire application is a huge WTF and if I had enough time I would re-write it from scratch. It makes use of .NET Remoting, which is a great thing if you do it right. However, my co-worker didn't quite grasp the concept...
Normally you'd have an interface exposed by a service and the client would call the exposed methods. My co-worker must have considered the process of creating a proper interface too time consuming, so he invented an interface which consists of exactly ONE method: "ExecuteCommand". This method takes a string containing the command, then a DataSet (!!) reference for the result and an array of string (!!) parameters for the command to be executed. It also returns an int value that informs about success/failure.
The service implements this method in the form of a huge switch-statement that executes methods (of a separate class containing only these methods - does that ring a bell?) depending on the command string. There are about 50 commands handled by this switch statement. Every case block contains a return statement for the success/failure-result. So there are about 50 return statements saying ... (*drumroll*) ...
return (int)0;