I only really have an issue with #2 and #6
8 isn't stated well - I think they point they're trying to make is "you can build concurrent systems
which make full use of your resources in python" - but yeah, python is not the correct language if you want to do more "traditional" concurrency etc - you do need very major architectural changes to the system to make use of those resources (e.g. re-architect to use message passing etc).
Honestly: I've never come across a situation where python's concurrency support was what let me down though ..
- if you're doing concurrent processing to avoid resource contention then it's support is fine.
- If you're doing parallel work for performance in a non trivial problem then you're only conceptually looking at ~ a 5x performance improvement available from concurrency anyway (on a 2x 8-core machine, with the extra overhead for increased fetches over NUMA, reduced cache efficiency etc). You can relatively trivially get a 25x performance boost for most of these kinds of problems (if they're data structure or numerical kinds of problems) by fairly trivial C extensions, and if you need more than that then I think you should really be re-architecting your solution..