You don't say whether you've recommended the papers for acceptance or rejection.
If you've recommended them for rejection (but nevertheless you view the work as being a reasonable specimen of the form, i.e., not egregiously flawed or crankish), then I think your idea to spread the opinions around is a good one. Sometimes I get sent a paper for which I think "I don't value this paper much, but so far as I know someone else really could." In that situation I often decline to referee it.
If you've recommended both of them for acceptance: well, you would think the same argument should apply, but I don't find it as compelling! Rather than declining to referee, you could just continue to read carefully and not assume the work is good because you found the prior work to be good. And if you really have a clear idea that the author is consistently doing good work -- there's no conflict of interest there! By consistently recommending their work for publication, you are helping them out but also the field.
Is “I’ve already reviewed many papers from the same author recently” a valid reason to refuse?
I think so, yes. In fact almost anything is a valid reason to refuse in the current climate: so many papers are being written and processed at this point in time that anyone person has to limit their involvement. Editors are getting turned down by referees all the time. If you decline, do it quickly and suggest other possible referees, in all my experience they won't bat an eye. As to
Is there any good way to decide how many is too many?
I think it has to come down to your own judgment, because (of course, right?) this also depends on your other professional responsibilities. As has been discussed on this site before, try to apply golden-rule considerations to your general refereeing policy.
Having said all that, I did want to make one counterargument: in all parts of mathematics in which I am familiar (which ranges from very theoretical to very moderately applied), refereeing papers takes a long time in part because the task truly is very intellectually demanding. For a given paper, there are a number of people whose expertise is so unusually close to the topic of the paper that they can evaluate it much more easily and quickly than most other experts in the same area. In many parts of mathematics, this number is very small -- sometimes, unfortunately, it is zero. If you are part of a cohort that has unusual facility and ease with the author's work, then maybe you can do in a day or week what some other (expert) mathematician would take months to do. In that case you should think again before declining, it seems to me.