Although this could be due to the "publish or perish" mentality, that often forces researchers to break down their work in several publications of lesser impact than make a single publication of larger impact, the fact that the "lifetime" of publications is getting shorter may also mean that the research is speeding up. Knowledge moves faster from papers, then to books, and then to being "common", and before you know it you don't really have to cite someone every freaking time anymore because everyone knows what you're talking about (I'm talking about things that are considered "common knowledge" here; you surely don't cite Newton every time you mention that white light can be broken up using a prism). More commonly, somebody will sum the "state of the art" into a book or in a good introductory chapter of a doctoral dissertation and people will cite that, instead of all the papers. Also, books keep getting cited for decades after their publication, so maybe a follow-up study could check whether there is a similar trend in the citation of books?
While the plurality of journals has made publishing quite easy nowadays, I don't think this is the reason for the observation that papers get forgotten faster. A bad paper will not even get noticed and will probably get cited only by its own authors in subsequent publications. Since we are talking about papers that do get cited here, this means that they have managed to attract some attention, and can therefore not be too crappy.