If you used CSS analyzer tools e.g. CSS Analyzer by Project Wallace you may have noticed the relation between selector duplication and declaration duplication.

Let's say we want to apply a padding of 1rem to three different selectors then we have multiple choices:

Choice one: Repeat the declaration for all selectors

.selector1 {
  /* other declarations */
  padding: 1rem;
}

.selector2 {
  /* other declarations */
  padding: 1rem;
}

.selector3 {
  padding: 1rem;
  /* other declarations */
}

Result: Higher declaration duplication score and higher declarations per ruleset.

Choice two: Combine selectors that share the same declaration

.selector1,
.selector2,
.selector3 {
  padding: 1rem;
}

.selector1 {
  /* other declarations */
}

.selector2 {
  /* other declarations */
}

.selector3 {
  /* other declarations */
}

Result: lower declaration duplication but higher selector duplication.

Choice three: Using of utility classes

.padding-1rem {
  padding: 1rem;
}

.selector1 {
  /* other declarations */
}

.selector2 {
  /* other declarations */
}

.selector3 {
  /* other declarations */
}

Result: lower score of both selector and declaration duplications but it violates the Separation of concerns concept and we end up with markup like this:

<p class="padding-1rem margin-0 text-bold text-center etc">...</p>

which is not really good for maintainability.
So, as an experienced developer how you deal with this dilemma?