Having some insight into the actual products and mechanisms by which spam filtering is done, I can say fairly conclusively that the biggest issue at hand is that consumers cannot and will not pay for the hardware necessary to do spam filtering anymore. Let me explain:
In days of yore, spam filtering was done with an appliance (or server/service) that did heuristics scanning. This mechanism worked well, but people found that there were a lot of repeat offenders -- so let's just block them entirely with an RBL (realtime block list). This is easy to do with IPv4, as we have a relatively small amount of offending addresses. The amount of resources to load this RBL into memory and process emails in real-time was reasonable.
Enter the modern age:
Spam emails have gotten incredibly sophisticated. I've noticed that some blatant spam has even gotten through gmail's filters to my inbox lately. Heuristics are a large part of it, but when they're so complex, we really need to lean even more heavily on RBLs. Unfortunately, that's no longer feasible. Now that IPv6 has taken off, if you end up on the block list, you simply change your IPv6 address, and you're back in business. Since we have (for all intents and purposes) a limitless supply of IPv6 addresses, it's unreasonable to have an IPv6 RBL -- especially one that *any* appliance can load into memory and process in real-time -- for the same cost as an appliance that would previously handle an IPv4 RBL.
Here, we find our conundrum: spam has gotten to the point of bypassing all but the most brilliant of heuristics scanning mechanisms, and RBLs are no longer an option. In order to have an appliance capable of handling *everything* that a modern spammer can throw at you, you'd be looking at spending tens of thousands of dollars of raw material cost on a server -- plus the software and overhead to develop the solution into a consumer-facing product. That is in contrast to the relatively inexpensive solutions that were so prolific only a few years ago.
tl;dr: IPv6 killed the spam filter.