Running pt-online-schema-change against big tables is pretty expensive, and in many cases, it turns out that existing --max-load and other settings are not enough to avoid problems. Often the parameters assumed when the tool is started turn out to be not optimal, and the overhead is too great to continue running. It may be annoying to have to interrupt the tool 80% of a 400 GB table was already altered, and you need to start from the beginning. 

In other cases, the replication topology may change while the tool is running, and correcting the list of replicas is needed to make the tool continue.

The idea is to have the interface for changing the pt-osc options dynamically to implement needed corrections instead of restarting the tool.