Better handling of cancelled scan / missing / incomplete datasets
If a scan is cancelled at the moment it might set some dataset (as rotation angle) to zeros because it has incoherent number of frame ...
A quick fix could be to take only a subset of those and / or append missing data to other dataset based on the expected number of frames like done in this patch nxtomomill.txt
But the correct solution would be to try to get all of the information contained. So for all first try to get all data, then if there is some missing frame create empty frames (or broken datasets to avoid creating extra data) with image_key set to invalid.
Then for other datasets complete missing data with a default value (zeros ??? )... in order to keep this coherent I would say that if at least one metadata / data is missing for a frame them just tag it as invalid.