Maybe their models already take into account this and the cost of lost commissions in such events is lower than the cost of increased scams.
"Now, should we tweak our process ? Take the number of merchants in the marketplace, A, multiply by the probably rate of false positive, B, multiply by the average cost of lost commissions C. A times B times C equals X. If X is less than the cost in increased scams, we won't do it".
"Are there a lot of these kind of issues?"
"You wouldn't believe"
"Which cloud company do you work for?"
"A major one"