At the time of writing, CellProfiler on the cluster will attempt to find images using paths to images from the client machine, rather than their actual paths on the cluster. The CellProfiler documentation on using CellProfiler from the command line indicates that the currently passed input argument -i images in the run script is being overridden by paths provided by Batch_data.h5. This file contains information from the pipeline which is necessary for loading analysis modules. The ideal fix is to adjust the pipeline before Batch_data.h5 is written with corrected cluster paths.