If it's kind of a risk, and you can't take any chances...Why are you testing in that environment?
Why not set up a VM with a test database, and a VM with a pseudo-cluster, and load a subset of your data, and experiment in a development environment so that you can know for sure - even if someone guarantees you the answer on here, you can not be certain everything is identical across all the versions of Sqoop, Hadoop, etc for them as it would be for you...if the data you are working with has value, you should find a safe way to experiment rather than trust your valuable data to the mailing list answers.
Now, in answer to your question:
According to my peer (I am not the Sqoop person where I work) if your incremental split is on a column that has increasing values, you can safely split on that, but if the value you split on is always the same, it is a bad choice for incremental splitting - he uses a datetime column I believe, and then the import is from the last imported datetime value up to the current max. I am not sure if that helps your case, but it is my hope that you find it useful.