That's not an accurate description.

For example, it is common in sorting algorithms to do an n² algorithm like bubble sort when the list to sort is small (e.g. <50), whereas when it's larger, we do an n log n algorithm like merge sort.

The issue is that merge sort does recursion, which comes with some extra cost, so that an n² algorithm beats an n log n algorithm, provided n is small.

It has nothing with your criticism to do.

You can (and probably should) add a threshold to recursive algorithms like mergesort so that they don't end up doing recursive calls on very small arrays.

For arrays smaller than the threshold, insertion is faster than bubble sort. And if you really don't want recursion for large arrays to begin with, you have heapsort or Shell sort.

I don't think there is any practical case where using bubble sort makes sense, except "efficiency doesn't matter for my use case and this is the code I already have".