There are a few cases found by a tool that are possibly over-catch of exceptions, especially those that will abort the server. Over-catching these exceptions may unexpectedly abort the server, and may cause problems in the future when code in the try-block evolves. I am attaching a patch against trunk that constrains the catch blocks to the exact exceptions that were thrown.
My tool actually found one more case in 0.96.1, but I found it has already been fixed in trunk:
Any feedbacks are much appreciated!