The material states
The main problem with this approach is that it can lead to unbalanced DB servers. For example, we decide to put all URLs starting with the letter ‘E’ into a DB partition, but later we realize that we have too many URLs that start with the letter ‘E.’
However, because they keygen is generating randomly, it is likely that we have a relatively even distribution of URLs starting with ‘E’ just as much as ‘x’ or ‘S’ or whatever other character you can think of. So this should not be a problem as it is quite uniform?