We are using SiftingAppender to direct logs for each job in our application to a separate file. These files are typically only written to a few times in pretty quick succession and then they're done. While the appender works well, we have found that we are creating jobs at a faster rate than it closes the sub-appenders, causing us to eventually run out of file descriptors. I know about AppenderTracker and will implement a solution using this, but this feels like this timeout value is something that should be configurable on the SiftingAppender. Or maybe just some kind of boolean to tell it to close sub-appenders more aggressively.