The way I'd look at it:
1) The risk of returns is well known. Easy to gather data on what proportion of goods are returned and at what time they are returned.
2) Therefore introduce a time lag to account for this between purchase and points crediting (IIRC, 4-5 weeks, which gives plenty of time for people to return unless they are absent or there are delivery delays).
3) Factor in some "breakage" where goods are returned after the points are credited, but allow for this. Again data on time of goods return could be used to factor this in
4) Set up flags for "unusual" transactions, ie. set up some parameters that would flag people with high proportions of returns (outside certain std deviations) after points have been credited, and flag these for further investigation.
I can't see, other than delay crediting for a few more months, another way of dealing with this. I suspect QF's system are working - not failing them - if they flag potentially systematic deliberate refunding after points are credited. The absence of wide spread reporting of this same type of issue (suspension of accounts) by other AFF members (after all, most on here are quite obsessed with accruing points) and the reluctance of the OP to engage us with further details on the proportion of returns and size of transactions, would suggest QF's flagging system is working exactly as intended.