The error should only happen in more than 50 connections are taken from the pool. The pool is indeed big. Is this easily reproducible? Does it happen when you disable the logger?
Could you provide more details, source code, etc., so we can check better what's going on (you can send to e-mail wagner at tmssoftware com)