1. Skimming the paper, it only matters if you need an efficient stable sort. If you just need O(1) memory you can stay with heapsort, which at least will have more reference implementations.

This lead me to look up browser sort implementations; http://stackoverflow.com/questions/234683/javascript-array-s... - it seems Moz uses mergesort and Webkit may or may not do something silly for non contiguous arrays.

So, there could be a use for it. For most applications you're about fine as it is.

2. I'm interested in hearing about applications where you're loading millions of array elements in people's browsers.

I was going to be cranky and make rude comments but I can envision people wanting to play with their data without loading it in specialized toolsets/learn R/build a DSL in $lang_of_choice.

I do some ML in browser for ease of visualization/portability. Don't often need to sort all the connection weights, but hey you asked :)

Here's my work-in-progress visualization with a dataset of 1 million+ IMDB entries to be stored in IDB http://dashdb.com/#/

It purposefully pushes IDB way further than it should be taken in most cases.

Crossfilter's link is broken (parenthesis and semicolon were included).