Ninja
Yes I understand how it works... and I don't like it for many reasons.
1)
Try to indexing a medium size database of articles.
I tried it on my site (20k articles) and JA Megafilter indexing has stucked on a blank page (I already open a ticket a few days ago). Now I'm trying to split all the articles with several filters and pray it doesn't stuck on indexing (who know how many articles are needed before stuck happens? How many minutes are necessary for indexing of thousand of articles?).
2)
In the front-end JA Megafilter at first load ALL the articles in the index and after apply on the client a filter to hide articles don't satisfied search criterials. In this approach there are at least two problems: loading time, memory occupation on the client (load all json items and render it in html with javascript).
Pagination is the best solution for faster operation without overload the server, but JA Megafilter's pagination it is dummy (because at first all articles are loaded and after filter on the client). In this case the pagination doen't exists.
3) Redundancy: all articles attributes and properties are already in the database... why recreate they and taking up space on server? An example over all (cause the size in bytes) are the pictures thumbnails.
This is my opinion (and suggestions too). Hope these can be useful for next release.