You keep saying “statistically impossible” without presenting any statistics. Different groups of voters return ballots at different times. If Raman’s supporters were younger, more progressive, more urban, or simply more likely to return ballots later, then later-counted ballots would be expected to look different from earlier-counted ballots. The fact that a later batch differs from an earlier batch is not evidence of fraud. To claim fraud, you would need to show that the ballots were invalid—not merely that they produced a result you didn’t expect. “That seems unlikely to me” is not a statistical analysis.