I think this is part of my argument: as datasets grow faster than single-core CPU speed, performance matters more and more.

It's hard to tell ahead of time if performance matters. When you hack something out, and are able to get it to work, it inevitably starts to grow features. At some point (around 1000 lines in my experience), Python no longer is fast enough, but the code is difficult to port to something faster.

It would be much better if you didn't have to back track on all the code you wrote. I've been burned by this enough times, that now I am wary of starting anything in Python, because I know it will grow to be bigger, and I will regret having picked Python!

Exactly. For web dev stuff, the database is usually the bottleneck. Caching is the next lace to put your efforts. I think I have had one problem where Python was too slow.

I'm with you here. I tend to use Python, but if I need raw speed/concurrency I'll reach for something else. Currently Go.