artificial intelligence: choice of the heuristic for the algorithm A * where the cost is less than the absolute distance

Looking for information on how to choose a heuristic for cases in which the cost of traversing an advantage may be less than one.

For example, let's say that movement is allowed in the cardinal directions. If all edge costs were at least one, then consider the diagonal distance. However, if the costs can be less than one, this could lead to an instance where the diagonal distance will be an overestimate and, therefore, is not an admissible heuristic.

My thought about it is perhaps to find the lowest possible cost and find a constant to divide the diagonal distance to ensure that it never overestimates.

I would appreciate some guidance or some resources to deepen this.