In a sharded cluster, dropDatabase (and the component collection drops) appears to occur in serial.

When databases, or the number of collections they contain are sufficiently large, the dropDatabase may not return quickly. This means that the time spent waiting will be exacerbated by the number of shards.