If you are combining thousands of viral genomes into the same index you may have some performance issues due to the large number of similar sequences (ie. all of the papillomaviridae). This may also cause a lot of issues with repeat sequence alignments.

You may have better luck pre-mapping the reads to the human genome then aligning the remaining reads to the viral genomes using megablast etc.

There may still be some issues using this strategy such as read pairs derived from an integrated viral genome where one read aligns to the human genome and another to a viral sequence. So you may have to write a custom solution to try and resolve these kinds of alignments.