How long to tell a good stock from a flat one?

Two analysts watch the same stock, which follows a GBM with volatility $\sigma = 20\%$. Analyst A says its drift is $\mu = 10\%$. Analyst B says the drift is $0$. You observe the whole price path continuously on $[0, T]$.

(a) Use Girsanov to write the likelihood ratio $d\mathbb P_A/d\mathbb P_B$ in terms of the path. Which part of the path does it depend on?

(b) Suppose A is right. What is the probability that the likelihood ratio favours A after 10 years? How many years of data do you need for that probability to reach 95%?

(c) Analyst C agrees with B about the drift but says $\sigma = 25\%$. Why does one day of continuous data settle the argument between B and C?