Updating priors given model re-run

At 1:36:25 of the following Youtube link (a speech Michael Betancourt gave in 2016),

The questioner asks about how to update prior information of some existing model when new data comes in hourly or daily. My question is a bit of a riff on this: if we have a model that is re-estimated intermittently (or that we have to use again when new data arrive), what is an appropriate time between estimations where we can consider new data collected as non-sequential? Stated differently, when can we consider data we collect to estimate a previous model as “new” and do we need to adjust anything about our previous priors in this circumstance?