Five to seven days, untouched. Size the daily budget so that the adset reaches three to five times your target cost per acquisition across that window rather than inside the first two days.
The window matters because performance varies by day. Monday through Wednesday does not look like Friday through Sunday, and an adset judged on three days is being judged on a sample that excludes most of a week. Facebook's own guidance runs longer still, closer to two full weeks without changes, which is probably further than most advertisers can stomach, but the direction is right.
This is really a budget instruction rather than a patience instruction. If your budget is set so high that you burn through five times target in two days, you have forced yourself into a decision before the data exists. Working backward from the threshold to the daily number is what makes the waiting possible.
Longer sales cycles need more room again. On a high-ticket offer where people take weeks to close, take the spend threshold, divide it across seven days, and then run at roughly half that daily figure, so you are not requiring the sales cycle to resolve faster than it actually does.