People use Python to actually do stuff. Performance-wise it's a nightmare but usability outweighs it in many many cases.
Often it's so fast that it doesn't matter at all that it can be 1/100 slower then C. 0.1 sec is still good even if it's not 0.001 secs.
Further, it doesn't help with the required restructuring; when the functionality is scattered across a dozen or more modules (complex software + good Python practice), then you first have to pull the functionality needed for the path you're trying to get fast together. That leads to duplication, removing abstractions and most certainly vetoes from other people.