How slow does an algorithm have to be before it stops saving time on level design, realistically? Obviously if it straight up fails for complex sets of constrai

How slow does an algorithm have to be before it stops saving time on level design, realistically? Obviously if it straight up fails for complex sets of constraints that’s one thing but even running for a day, if it yields usable results that seems like a timesaver.