The author raises some good points.

However scaling across multiple workers while using a reverse proxy cache like nginx isn't terribly hard. And it can help make up for a less concurrent language/framework.

Load balancing is often necessary, but the necessity can often mask other underlying technical debt.

Parallelism through multiple native processes can have a high cost from process overhead, effectiveness of IPC style chosen and separation of duties.