I'm not sure what trends you are looking at, but most consumer startups usually find it unnecessary to look at historical data points past X months. Things (features, users, etc.) change so much that averaging in those data points so far back would misguide you at the present. I would go with a solution that is almost perfect for you if you started today (for example, Mixpanel might be it) -- gives you the best odds for your future analysis. During the couple weeks after switchover, I would just export the Mixpanel data to a local spreadsheet (and run in R as you mentioned) with your other data. After sometime, you won't need it.
However, I don't want to reinvent the wheel, or to invest a lot of time in something that other apps like Mixpanel, NuConomy already do very well. I suspect Mixpanel uses the timestamps of the events for billing and if so they might require some non-trivial changes to their business logic to allow historical data. I hope they will, someday. Until then I might settle on sending them only new events and using R on the historical data.