Custom Template model training fails with more than 8 selectionMark columns in a table

Question

I am using Azure AI Document Intelligence Studio to create a Custom Template model for my documents. The model should be able to recognise a table field on the document, in which I have defined 21 columns of various subtypes. I wanted to define 10 of them to be of the type selectionMark.

As training data I have 6 .pdf documents and the Training process succesfully completes fine only when I choose to have only 8 (!) of the desired 10 columns formatted as selectionMark which is very unfortunate.

The documentation doesn't mention anything about this limitation and unfortunately the Error Message when the model fails is very generic:

Model training failure

InternalServerError: An unexpected error occurred.

My questions:

1. Are there limitations as to how many columns in a table field can be chosen to be selectionMark? If not, what is needed in my case to let the training process successfully complete?
2. Where can I inspect any logs for my Document Intelligence Resource's model training process to find out the root cause of model failure? So far I have spend a lot of time to create minimal working examples with combinations of labeling for my data, which doesn't scale resources-wise.

I am using API Version 2023-07-31 (3.1 General Availability)

Thanks a lot!