I'm having trouble wrapping my head around how that would even work. I get scaling, it's a common enough thing but:
1) Mixing scales like that should just produce garbage.
2) 50 digits is a *lot* of digits. I'm intensely interested in what could possibly require that kind of (scaled) precision?
The other "large" data type in Progress is int64. But according to Wikipedia the largest unsigned value for int64 would be 18,446,744,073,709,551,615 which is only 20 digits.
I think your customer is going to have to come to grips with reality and use unscaled numbers.
(I think I remember from college Physics that you can model the entire freaking universe down to the Plank length and you only need pi out to about a dozen digits. Or something like that...)