If the bacterium could only acquire many fragments at once, it would be far more difficult - perhaps even impossible - for the phage to acquire enough mutations in the specific targeted sequences to escape. Suppose a phage was targeted with Cas9 at multiple sites. Most phage in any given generation will have only a handful of mutations in the entire genome. It's enough to ensure that the phage population can escape targeting at a single site, and usually at two sites. But three is much less likely unless the phage population is very large, and four still less likely. Escape becomes exponentially less likely with every additional targeted site.
Here's a simple example. In nature, the CRISPR systems that use Cas9 work by acquiring fragments of invading DNA from phages - viruses that infect bacteria - or parasites and direct Cas9 to cut them. This gives the bacterium resistance to any future invader containing that particular sequence: CRISPR is an acquired immune system. But these fragments are normally incorporated one or very occasionally two at a time, which means it's not difficult for phages to mutate the targeted sequences and evade cutting. This makes it an endless arms race: bacterium acquires new fragment, phage mutates to escape.
As an enzyme that can be directed to bind and cut almost any sequence, the Cas9 nuclease has revolutionized genome editing and regulation. But its potential for sculpting evolution hasn't yet been fully appreciated, much less harnessed. Simply put, Cas9 privileges design over evolution because it lets us alter genomes in ways that are hard for evolution to undo.