...
However, PhixFlow starts lining columns up from the left, based on the idea that if there are missing columns, they will be missing from the right hand side. This means if columns are missing from somewhere in the middle or at the beginning of the line, the line may fail to import if the data types don't line up. However, this is not an error trap, it is essentially arbitrary - if you think lines might be missing values in the middle or at the beginning of lines you should be careful about using this option – in any case, you will need further validation.
...
If you set a number in Errors Before Rollback – providing that, across the run (not individual files), the error count is less than the threshold you’ve you have set - this will load all records that are not in error. You will get a warning message about any that fail in the console – with the file name, and details of the line that failed. But these are essentially discarded from the import. These files will then be placed in the archive directory.
If, across the run, the threshold is reached – all files that have been processed up to that point will be placed in the error directory, including files don’t contain errors (because these are associated with a failed run). The run will be in error - that is, a red line in the console, the resulting data set for the run (the “stream set” stream set) will be invalid, and therefore not available to the next step in your process. Any files remaining after the error threshold was reached will remain, untouched, in the input directory.
...