Hello tazzo,
Originally split-in-chunks script has been designed to process large projects to a single mesh model, that couldn't be done in a single chunk due to the algorithm limitations in the older version (it existed for dense cloud based mesh in Arbitrary mode), but now most major tasks are optimized to be run in out-of-core mode to fit 32 GB RAM.
So it should be possible to build a dense cloud or mesh in a single chunk for a dataset consisting of thousands and tens of thousands of images.
Is there any specific purpose for your projects to use the script?
Maybe you can also share the screenshots demonstrating the issues of the merged model in comparison to the model processed in the single chunk.