Unless the two choices are extremely close, we probably don't want it to be the case that a single investor, even with 100% of their funds, can single-handedly change the outcome. We just want each investor to be incentivized to put their money behind what they believe to be the best policy, and hope that collectively they choose right.
Thus, we should expect that even if our oracle knows with certainty that A is the best policy, and invests all their money into it, they're unlikely to be able to unilaterally change which policy will actually be implemented. They will therefore be incentivized to bet on B succeeding (which is also a winning bet), and now that we're revealing that they bet 100% of their capital on it, all we've done is magnified the signal of that bet. But this is bad - our oracle is betting in a way that moves us away from the policy they know is best.
Thus, we should expect that even if our oracle knows with certainty that A is the best policy, and invests all their money into it, they're unlikely to be able to unilaterally change which policy will actually be implemented. They will therefore be incentivized to bet on B succeeding (which is also a winning bet), and now that we're revealing that they bet 100% of their capital on it, all we've done is magnified the signal of that bet. But this is bad - our oracle is betting in a way that moves us away from the policy they know is best.