In high-reward / low-risk environment, such as building an indie turn-based retro-style game, go with your gut feeling unless you have a good reason not to.
In a high-risk / dubious-reward environment, such as implementing cryptography, follow the best practices to a t, unless you know intimately how things work and maybe codified some of the practices.
There is a wide gamut between these two extremes.
In my experience, many "best practices" are the pitfalls you should be wary about, as they can easily translate into hundreds or thousands of lost hours of work and derail and doom entire projects. (The most annoying part of this is that the real causes won't be found, precisely because "best practices have been followed". Therefore the reputation of the best practice will stay untarnished).
Cryptography on the other hand is a well known example of something you should not touch at all unless you are an absolute expert- that's not even a "best practice" but probably the only reasonable practice.