Hi @reige012,
It sounds like concatenation (or some other modification) is causing both problems.
This indicates that all but 5071 sequences are shorter than that length. This appears at all bases > 1 nt, so something is definitely wrong.
We have only seen this error one other time, and that was similarly when some sort of pre-importing modification was being used (though that was never clarified nor resolved): Quality-filter paired end illumina
I agree, concatenation should not be a big issue, but clearly something is going wrong — make sure you do not have blank lines or special characters in your sequences. In the end, using a manifest format to import your sequences and grouping duplicates later on may just be a more hassle-free approach... while I would personally concatenate the sequences before importing, it is not worth it if it makes QIIME 2 unhappy! The more general rule I preach and observe is to not modify my data prior to importing, to avoid headaches downstream.