This paper reports a practical experience with memory analysis on a real world complex middleware platform, being developed in the context of an academic-industrial collaboration. The reported experience suggests a practical method that can help practitioners to analyze memory leaks and to adopt proper actions to mitigate these bugs, especially in the context of complex Off-The-Shelf based software systems, and in some cases it highlights issues still open. Indeed, we experience that fixing a memory leak, when possible, might be not enough to solve the memory exhaustion problem.