I send time and three temperature readings to SD at ten second intervals, i.e. 24 items/minute, and that gives about 260k/day. If the shape of your data is similar, this implies you would send about 1Mb a day, which makes a cheapo 2Gb SD card look a pretty good prospect.
I don't see why a Mega can't send 1500 values at a time. It's just a matter of reading off the card. I think the real problem is finding the time to do it, as there are too many "could be"s in your post, and it might be simpler to pass them on as soon as possible, rather than accumulate anything.
I believe you are obliged to use two Arduinos because you need the first as a data buffer. This is due to the nature and unpredictability of the data, but the problem is that the second is also a data buffer - just bigger and slower.
The nature of the data, if you can call it that, is unclear. If the input is just a pulse, that isn't data. The data is the measurement - when, and for how long, which is done by Arduino. This is starting to look like a good time to learn up about interrupts, of which the Mega has several. You might find you can get by with one Arduino.