Even if the sign detection bug were somehow fixed, the direct usage of exp(10, N) in EVM for \(N \approx 76\) is still borderline for 256bit arithmetic. In particular:
Thus, 10^77 in 256bit arithmetic would overflow to \(\mod 2^{256}\) and produce an incorrect result.
