Select files depending on their versions

I have made good progress, but I am encountering a difficulty for which I have no solution.

I have created a managed folder where my files are in HDFS. This allows me to manipulate them via the API. Then for each of the subfolders (these are YYYY-MM partitions), I run the following query to create a copy of my latest version of the file (here by sorting down the file name) whose name ends with "_latest":

import dataiku
import ntpath

dossier = dataiku.Folder("zzZZzzZ")

for partitionn in dossier.list_partitions():
list_files_in_partitions = dossier.list_paths_in_partition(partition=partitionn)
list_files_in_partitions.sort(reverse=True)
with dossier.get_download_stream(list_files_in_partitions[0]) as f:
dossier.upload_stream(ntpath.split(list_files_in_partitions[0])[0]+"/"+"Chiffres"+str(partitionn)+"_latest.txt", f)

On the Notebook, the code works very well. Then I use "Files from folder" with the following regex: ^.*?latest.txt$ to filter only the _latest created in each subfolder.

Then to launch the code each time a new file arrives in General, I create a trigger "Trigger on dataset change" and select the managed folder.
Everything works fine.

The problem is when I put my python code in the step of my scenario and it runs, I get the following message:

java.io.EOFException
EOFException

I am aware that this does not give much information but do you have any idea anyway?

Thanks

Ineedi2