I suppose this is close to my heart because I just worked on it, but if you chop the snippet on the client side it defeats the logic used to pick the "best snippet".  That logic isn't that great in Cirrus now but it'll get a whole lot better when we deploy the new highlighter.  Right now the snippet is always 150 characters or something +/- 20ish characters on each side to find a word break.  We pick the best snippet based on hits in the 150 character window.  At minimum we should let you configure it to something that'll fit better.  I suppose the best option would be to configure up font widths that matter and then use them to chop really accurately.  For the most part that sounds pretty simple and quick to implement so long as we're ok with estimates that ignore stuff like ligatures.

Do the type timers fire the hit on the leading character or on a slight hesitation?  Prefix search on the site is leading character and then it <cancels> the request if the user types more.  That is silly because I can't cancel the request on the backend....  If it triggers on a hesitation then we should just plow ahead, I think.  If it triggers on leading characters then we should totally cache requests shorter then N character.  3 or 5 or something.

From the code:
        // If we were constraining the namespace set, we would probably use
        // 0|1|2|3|4|5|6|7|12|13|14|15|100|101|108|109|446|447
        // to keep it related more to article, article talk, policy,
        // policy talk, help, and help talk types of resources.
        // The odd numbered Talk pages could even be witheld, but that's
        // sort of pointless when the number of backlinks to them is
        // likely to be small, meaning they won't turn up too much
        // unless they're a result(set) of last resort, or the user
        // went to the trouble of prefix namespace searching such as Talk:Cats.
        // But realistically, it's probably easier to just stick to
        // not defining a namespace constrainint set, and thereby (likely)
        // getting more pre-cached responses, due to other consumers leading
        // or follwing suit. There's a school of thought, or there could be,
        // that says only namespace 0 should be searched here, as it's
        // the core article content. But users may practically want
        // categories, too. And such logic spirals out from there.
        // If we were instead using the opensearch API and were seeking
        // parity with the desktop and mobile web experience, we should
        // indeed as of 27-March-2014 only be searching namespace 0.
        // But as CirrusSearch will be the norm and server load is expected
        // to handle things just fine (no fallback is necessary per Search team),
        // higher quality search results now can be obtained anyway.

Cirrus searches all wgContentNamespaces by default and it is optimized to do so.  All non-content namespaces are in another index so we don't have to pay attention to it during the request.  We also don't have to filter by namespace at all.

Each namespace has a weight factor that influences its position.  That factor often ends up being more important then links.  Links are "score * log(incoming_links + 2)" and the weights vary from "score * 1" (MAIN) to "score * 0.0025" (TEMPLATE_TALK).  Our power users expect these because lsearchd did it.  Mobile users, who knows.

        // With all of this considered, we want a request of the following format
        // //en.m.wikipedia.org/w/api.php?action=query&list=search&srsearch=cats&srprop=snippet|sectiontitle&srlimit=15&srbackend=CirrusSearch&format=json
        // Note that MobileFrontend's use of opensearch has its result
        // set limited at 15. Note also that the 'srprop' only keeps 'snippet',
        // and 'sectiontitle' plus the 'title' field which is always implicit.
        // This buys us some additional features once we're ready for them,
        // all the while populating the cache.
        // We probably also will want to add 'srinterwiki=1' in some future
        // state so that users don't have to change their language to
        // search setting. As it is, 'srinterwiki' is not yet in place
        // and the format of such results may look a bit different,
        // so it's probably best to hold off on 'srinterwiki=1'. We are
        // not yet using the snippets and section titles, but let's get the
        // cache populated for our sake and everyone else's sake.

Interwiki is coming but I'd give it a few months, I think.

        // NOTE:
        // Although as of 27-March-2014 it seems that suggestions may not be coming
        // back for CirrusSearch as frequently as for Lucene, that's probably
        // just an artifact of relatively lower training of suggestions.
        // In other words, it's likely that the suggestion pairing will grow.
        // Currently, we're not examining [@"query"][@"searchinfo"][@"suggestion],
        // but we could. There are two cases for the suggestion.
        // 1. When the result set is of length 0, just fire off a search with the suggestion.
        //    This is the case where the user probably mis-spelled something.
        // 2. When the result set is of short length (less than 5?), fire another search with
        //    the suggestion, and then collate those search results /after/ the first result set.

The suggestion is actually better then you give it credit for: even if it lots of results show up if we provide a suggestion it might useful.  It comes from redirect and title names and it'll suggest combinations that work.  So if the user searches for "picket's charge" it'll suggest "pickett's charge" even though there are plenty of results for the first term.  The results for the second term are better.

The reason you get different results is because the implementations are vastly different.  The Cirrus implementation has less tuning but is "more modern".  Whatever that is worth.

Nik