Indeed. According to this paper even "the most carefully designed and thoroughly tested software on the market still averages more than one bug per 1000 lines of codes". And Peres' Law (see the paper) should inform how we deal with that. Testers should strive not only to eliminate bugs, but to see to it that software will report bugs when they occur in the field. This means the guidelines for Windows Error Reporting should be followed (and tested). I feel these guidelines should be a logo requirement.