Interesting article. Your final solution is indeed a very good choice.
However, using a Razer Blade for deep learning with Linux is not very complicated to set up and offers very good performance.
I have the 2019 Studio Edition using Pop Os 20.04. This OS is specifically designed for productivity. It is very straight forward to install all needed packages for Tensorflow.
Setting up the Tensorflow Object Detection API for TF2 was significantly more straightforward than on Windows and comparable to MacOs.
Also I do not have any issues regarding the Razer Blade Laptop in a Linux environment. From what you wrote, it seems to be quite an advancement from Ubuntu 18.04. to 20.04.
I just want to make sure that people don't get the feeling that Linux based Laptops cannot be used for Deep Learning.
If it is a good choice regarding pricing and longevity is part of a different discussion.