Something that pops up pretty frequently is to implement time decay, especially where you have recursive chains of jobs. For instance, say you want to keep track of a popularity score. You calculate today’s output by reading yesterday’s output, discounting it by  $$ exp(-lambda Delta T) $$ and then adding some hit count for today.

Erik Bernhardsson

A technique for computing time-decayed popularity scores without storing timestamps. Instead of tracking when each score was calculated, you factor out the current time as a constant multiplier and store only the time-independent component. Because the raw accumulated value grows astronomically large, you store its logarithm instead. Adding new terms to a log-space sum is handled via a numerically stable log-sum-exp identity, avoiding overflow. This approach was used at Spotify for track popularity scoring.

A neat little trick with time decay