“Scale up”, a software term that has almost zero value when translated to aerospace engineering.
So you're saying that after say the initial 100 launches
Launch 100-199 would have 5 failures
Launch 200-299 would have 7 failures
Launch 300-399 would have 10 failures
Why?
Generally doing more of something reduces the cost per something. Why do you think the inverse will hold with spacex?
You need the demand to make that worthwhile though.