Hi @Siegfried Heintze ,
I realize since you've asked this question a while ago, you've probably explored various methods of doing this file processing. Just in case you're still looking for a path forward, instead of maintaining session state with user file uploads, storing the files on NTFS drives for processing, and keeping track of it all; what I suggest using is an app service web front end and an Azure function API back end.
Use Tutorial: Upload and analyze a file with Azure Functions and Blob Storage as a guide. With the app service, is just handing off the upload the request so you should be able to get away with a low resource S1 and function on a consumption plan; where you only pay for the compute.
Again, using the above tutorial as a guide, I would do the following:
Web App -> upload file to blob storage -> Azure function triggers off blob upload -> process the file for XML composition -> store the file in a separate blob container.
The beauty of the consumption plan is that it will fan out and back in based on demand. To keep things organized and nice and tidy, the web end would set the container name for the client request which you can then pull from the trigger request.