I would pay for this! It's common for blogs to go through a life cycle, and usually when you discover them they're already relative inactive. You can go through archives but there isn't good software support for making sure you see everything.

This would also be useful for going through historical incidents - e.g. replaying the top 10 politics blogs, day by day, during momentous events. It's simple enough to just treat it as an offset; the display would clearly say the original date of publication. This is a lot better than simply adding old RSS feeds since it comes in at the same rate it would have happened in the moment.

Something similar could be set up for newspapers. Imagine receiving all space-related stories from Life, NYT, Guardian, Spiegel, for the time period from say 1965-1970.

The current best version of this requires a huge amount of research into old newspapers, and also reading books and then manually connecting each book's timeline. If instead you could in parallel consume multiple sources, the correlation would be natural.

Even better would be adding in later commentary about those events.

Example: take "History of the Decline and Fall of the Roman Empire", and annotate each page with contemporary thought from each era. So every page would a section on what the scholarly response was when it came out, then 50 years later, then 100, each adding in new methods of investigation and validation/testing of the claims as archaeology, linguistics, carbon-dating, anthropology, etc. developed in the background.

There are some services like Feedly that cache everything some users did use, so you can go backward even if you subscribed just now, but unless someone usees it, I doubt it will work as it implies service must crawl internet for RSS feeds and subscribe to them all the time in order to keep history (since RSS on its own can't do that).

Sounds like internet archive for RSS.

I once made a Python commandline utility for this, which does still work if you can install it with the right (now long-outdated) Python version: https://pypi.org/project/dripfeed/

I don't intend to maintain it further but all source code is available and it's not terribly complicated (all the hard stuff is done by other python libraries).

Google reader used store every entry in the feed ever, and it was my primary reason for using it. I was subscribed to tons of webcomics, and this allowed me to easily keep track of where I was. The only limitations to this was the rate they checked feeds at (I never read about any complaints about this, but also wasn't looking and wouldn't have been interested at the time), they only had history starting when the first person added a feed, and a large number of blogs and comics would only put notifications in the feed, without any actual content.

Well if the feed doesn’t contain older entries, it’s impossible without something like the Wayback machine (and I doubt they store rss feeds)

Some feeds are just paginated. Following "next" links would work. Feeds readers don't do this though as they assume people are usually interested in the new posts.

Perhaps the tool behind https://rewind.website ?