Really like the idea in Julien’s patch, that could be very useful. Maybe some of the continuous data could be thinned so the colls don’t get too huge, simple enough (speedlim or something). Also you could route the separate objects (or groups of similar ones) to different colls. To track the mouse use mousestate, if you think that data would be useful too.
Continuous data can also be nicely seen with a multislider in scrolling mode, perhaps with several streams in one. You might also consider *mtr*, which could play back everything that was done at the speed it was done… also a nice way to re-do the performance itself! (32-track limit though, so maybe banks of them?) — With the mousestate data in your mtr, you could play back the mouse motions too — ;max pupdate $1 $2. Help! My mouse is possessed!
The data in the mtr could be parsed to generate the timestamps readily enough, with a bit of fiddling. The colls in Julien’s patch do it nicely… the mtr is just a different way to "revisit" the actions of the performance — seeing it as it happened rather than as lists of numbers, which might be illuminating in other ways.
A very interesting idea for tracking user interaction—I’m considering something similar for testing educational software (to see where users spend time, what they click the most, etc.) Graphical representations of the "usage hotspots" (similar to those used for website analysis) would also be straightforward — overlay "use" colors / pixels onto a screenshot of the patch, so areas gradually get more and more saturated as the particular objects get used more. That would be pretty interesting to see.