Add one 'if' condition to ensure the first file of each dataset exists. If not, metadata is assigned to NULL.

Bugfix for Start() parameter 'path_glob_permissive' and write the explanation in FAQ and Start() header.

Remove the indices_chunk part to the first, so if the array doesn't need reorganization, the new development is skipped.

Add additional data array rearrangement if merge_across_dim + split_multiselected_dims. It corrects the displaced values when the values in the same file are called more than 1 time.

Change final_dims_fake back after reoder array because if retrieve = F, it will be used in attributes later

Move adjusted split dim order part to after 'Detect dimension' print. Then reorder dimension back to the user-required order after getting array.

Add new parameter 'merge_across_dims_narm' and develop codes for the case 'merge_across_dims_narm = TRUE'.