I believe that this depends on the specific action to be achieved, at the end of the day what this instruction seeks to do is to try to keep track of the way in which the information is processed, leaving the logical sequence of how the result is reached. It may help but sometimes it is a cost-benefit issue.

A very long prompt, thinking in the case of an app, is an almost predetermined recurring expense.

I don't think it is luck, this concept is called "Zero-shot-prompting" 

it helps the LLM give better answers.