Hi cecki,
From the status messages we suspected the same as you, that it was caused by an IOException. Probably there are problems with disk (we are checking this) but the thing is that we were using log4j for 2 years now and working with no problems, and the day we switched to logback, we had this problem almost every hour (no fixed pattern here). We had do restart our application in order to get loggin back. Our disk was not full, so this is not the cause. We reverted back to log4j over slf4j, to get back logging without problems. Of course there are some differences between configuration from logback and log4j, so probably this configuration differences make it work without problem in log4j. I assume that the problem is an IOException on renaming the file that leaves the appender closed, so no more writing on this appender is possible. In our tests this didn't happen. we didn't test with as high througput as production environment, and also disks were different so the IOError might not have happened.