There are different strategies for numbering displayed math. The most common are

1. Number only the formulas you reference to. It makes your paper more clean and gives more freedom to the editor (i.e. making this math inline).

2. Number all displayed math. Even if you don't reference your formulas, take care of those who will read your paper and will want to reference them.

3. Number only very important formulas and the formulas you reference to.

I am thinking, which strategy to choose and I want to make an informed choice.

Which important strategies have I missed? What are pitfalls and benefits of the strategies, described above? How important are these benefits/pitfalls?

Ideas from the answers below: strategy 4. Check with the journal you are planning to publish your contribution.

The similar question was asked here. The most popular answer was "Use strategy 2: number all your displayed math to help the reader to reference your formulas". However it is not clear for me if it is ok to have a 40-page paper with 150 numbered formulas. It seems a bit crazy to see "it follows from (146)". Also it was noted, that this strategy violates "Checkov's gun principle", i.e. fills the paper with irrelevant details (numbers). However the question, stated there was a bit different. I think that that question was completely answered, but not this. For me it is still unclear, whether it is ok to have formula (146) (or formula (1353)) or not. The question made community wiki, so you are free to improve it.

"don't use too many indices, don't use nested indices/exponents, be careful when you number an equation, what if you want to insert new material before?") , you realize how amazingly freeing modern technology is. – Thierry Zell Oct 15 '10 at 14:48