What I understand of your question is that you want to load into a resident table the information coming from several historical QVDs with different structure; they donot have the same fields, right? If I am right look at CONCATENATE sentence (previous to LOAD sentence) in order to join into the same table information from different, but similar, sources.
Hope it serve!
I'm not sure if I understand your problem right but I think you need another approach by avoiding the set error bypass and loading the qvd's correctly from the beginning. I assume that you mean with "variable names" the included fieldnames of the qvd's and not the filenames of them and that your problem is that not all qvd's contain the same fields but you want to load them per wildcard-load without specifying special load-statements for each of them.
There are several ways possible to handle these situation. For example, are there quite a lot of files but only a few different data-structures you could use a single subfolder for each different type of them and using than a few specialized load-statements. Another method would be to concatenate all qvd's to a dummy-table:
load 'Dummy' as Dummy autogenerate 0;
load * from path\qvd-files*.qvd (qvd);
drop fields Dummy;
But of course you could also read all fieldnames from the files and create a load-statement from this on the fly. By qvd's it's quite easy because you could use a loop per QvdNoOfFields() over all fields and per QvdFieldName() you will get the fieldnames.
I mean that the use of the error-mode should be always the last measure if an error ocurred and there is really no way to implement the requirred logic without such a bypass to ignore the errors. Ok. that was generally spoken - in the real world it might be more practically respectively more pragmatic to choose those way which is easier to handle either avoiding the errors or correcting them afterwards.
In your case it would be more expensive to correct these errors if a quite simple forced concatenating would avoid it in beforehand.