For me, it starts during sprint planning. I know what work is going into the sprint, and what work is being assigned to me (we don't run sprints in the natural way of grabbing the next card that's free. I disagree with it, but it is what it is). So I plan out what I need to do on a day-to-day basis to finish that amount of work. For a normal sprint (i.e executing on plans and agreed upon designs) I try to plan the day to be ~6 hours of coding work a day. Sometimes my estimates are long and I finish in 4 or 5 hours. Could I start on the next card? Absolutely. And if I finish all my work early, I'll pull something from the backlog.

But in general, I find it more sustainable to just go home after I finish in 5 hours because I know there are going to be times that I work 10 or 12 hours (or a Saturday). Also, for me, depending on the next card, it might take an hour or so to get going at the end of a day. So I'm now at 7 to 8 hours which is when I normally leave. Was that hour or two wasted? No. I retain the knowledge till the next day. But again for me, I've found that coming in and starting a new task fresh reduces that one or two hours to less and gives me a singular focus for the day.

If I finish in 2 or 3 hours, I always move on to the next task. That would be silly to go home after 2 hours (most days).

I adapt this style for wherever I work. It has worked for me at start ups as well as BigCo.

[Update] For the side things (refactoring messy code, flakey tests, automating a tedious process, etc.) - I create cards/tasks for them and get them into the sprint.

I know this makes me sound like a "terrible" employee. But it allows me to give my best on a more sustainable basis. And if it is something that truly needs to be done in a month, I'll work those 10 or 12 hour 7 day work weeks...