Due to limited memory, I think most compression algos have a window of "stuff they've seen recently" which they use to detect duplicates. If the duplicate is of

Due to limited memory, I think most compression algos have a window of "stuff they've seen recently" which they use to detect duplicates. If the duplicate is of something too far back, it's fallen out of the window and been forgotten.