> You tell me crashing was expected default behaviour to a point where the suitability of implementation details was decided by how much one would crash? No, a

> You tell me crashing was expected default behaviour to a point where the suitability of implementation details was decided by how much one would crash? No, and Chen emphasizes this point repeatedly in his blog post. The customer knew the crashing was bad, but it wondered why switching to the unordered map caused the frequency to increase. For threading-related problems, an increase in the crash frequency can be a good thing, since it makes the problem somewhat easier to reproduce and debug.