"To overcome these self-inflicted challenges, DolphinDB stopped hiring noob developers trying to make HFT systems in Python, instead preferring run-of-the-mill C/C++ developers who understand heap vs stack allocation".
Edit: oh wow, check the poster's previous submissions :)
I love the open-ended nature of the title tho.
It doesn't say how an effective, efficient, nor industrial low-lantency engine works, nor what a microsecond level engine is (insofar as a label cutoff).
Honestly, boring heap/stack practices with mindfulness about the application workflow and memory management therein goes miles.