Thread.Sleep issue

Perhaps I misunderstand what you are doing with a semaphore, but when I say that the thread is suspended by Sleep, I think that it does not work the way you want it to. If the semaphore is in the thread, then the Sleep won't let it do anything during the Sleep.

You are certainly doing more in the catch block than I would. I don't know if it is stated explicitly in the documentation not to do that but I think it is worthwhile restructuring the program. I think it is best to treat exceptions as exceptional, not normal and you are using an IOException as normal processing. If possible, check for whatever condition that would result in an IOException before attempting the operation that is in the try block corresponding to the catch block. Another possibility is to set a flag or something in the catch block that the program uses later.

I certainly recommend using a timer instead of Sleep. The only advntage of Sleep is that it is easy; there is no other advantage. A big disadvantage is that it puts the entire thread asleep, so it will prevent the thread from dong anything else. That could be related to the problem but I am not sure.

I've attached the relevant code.  It basically monitors for IOException, writes the error message out, checks the error count, then eventually puts the thread to sleep for five minutes, then retries the process.  (my service oriented application works by continually monitoring and processing XML input files).  After the third retry, send out notification email and shut down the process.
When I do the thread.sleep(300000) in a different error condition, say invalid XML format (just as a test), the five minute delay works fine, leading me to believe the nature of the error being monitored for (IOException) inheritantly causes the problem.  I could either use a longer delay period, or try to use a timer function instead of thread.sleep, but I'm thinking that using a timer process for creating for the delay when monitoring for IOException may just have the same problem.

Any ideas?  Thanks!
******************
              catch (IOException e)
                {
                    StringBuilder sb = new StringBuilder();
                    sb.AppendLine("IO Exception thrown:");
                    sb.AppendLine("===================================");
                    sb.AppendLine("Message: ");
                    sb.AppendLine(e.Message);
                    sb.AppendLine();
                    sb.AppendLine("Stack Trace: ");
                    sb.AppendLine(e.StackTrace);
                    sb.AppendLine();
                    sb.AppendLine("Path: ");
                    sb.AppendLine(SOURCE_FILEPATH);
                    string[] to = new string[] { "scott.lackey@*********.com" };
                                       
                    Event error = new Event(5, e.Message, e.StackTrace, SERVICE_NAME, errordate);
                    EventProvider.Add(error);
                    error_count++;
                    if (error_count > 3)
                      {
                          
                          Utils.SendEmail(to, cc, "scott.lackey@****.com", "APPLICATION ERROR: RunQueuer", sb.ToString(), MailPriority.High);
                          Utils.StopWithErrorEntry(5, e.Message, e.StackTrace, SERVICE_NAME);
                      }                   
                    Thread.Sleep(300000);

               }
************************

I can't say much before I see the code, but I have a theory... You said you have the Sleep call in a catch block, that would mean that thread sleeps only when exception is thrown, so maybe those 30-60 seconds are there because of the normal flow of the program and exceptions are never thrown.
But it's unlikely that you have a code whose difference of execution time is that big, unless you're testing the code with different data each time...

Can you post that part of the code and any part related?