Scalene author here. I was not aware of either tool - many thanks for the pointers! I just tried py-spy (I will try PyFlame on a Linux box momentarily). It's pretty cool, though having to run it as root on OS X isn't great (scalene runs without the need for root privileges; the CPU profiling part is pure Python). Py-spy does appear to efficiently track CPU perf at the line granularity (and does a lot of other stuff). It does a lot of things that scalene does not do, but not memory profiling. Also, I personally prefer scalene's line-level display of annotated source code vs. the flame graphs, but YMMV.

The possibility to run py-spy on an already running Python program in production is pretty awesome. Do you think Scalene could do this?