Consolidate Timestamps and Time Windowed Dataflows

I've not quite got around to trying this stuff out but from reading the docs I suspect this may covered by the empty chapter found here:

https://timelydataflow.github.io/differential-dataflow/chapter_6/chapter_6.html

I have two main questions/problems I'm aiming solve.

Firstly, I'm unsure how to go about consolidating differences from higher resolution timestamps to a single, lower resolution difference. E.g. perhaps 5 separate differences occurred within a minute but I only care about knowing what happened within the minute, not exactly when within the minute. I think the same logic could apply for any kind of timestamp that was a numeric type.

Secondly, I'd like to have a dataflow retain changes and reverse them upon a specified time window passing. Essentially forming a temporal window.

In both cases, I think it would be solvable to have logic be applied to the data prior to inputting it into the dataflow. However, I'd quite like a reusable operator for each. I don't think I can achieve either outcome with existing operators but I could well be wrong. Does anyone have any ideas how this could be approached? Are operators for these outcomes going to result in some unforeseen issues?

Thanks for your time regardless of whether you can help!

Start with the empty Chapter 6 page linked in the issue and review the existing operator documentation and examples around it. Clarify whether reusable operators for timestamp consolidation and time-window reversal are feasible, then define their behavior and acceptance criteria before implementation.