Brahms also does this at the end of the Rhapsody No. 2 op 79, a 'written-out rit', most effective it is too...

I wonder whether the xylophone solo at the beginning of the third movement of Bartok's Music for Strings, Percussion and Celeste counts as a written out accelerando and then rit.

And on the subject of tied hemidemisemiquavers, I took delivery of a vocal score this morning, a new edition, and I noticed that all the quavers for the vocal lines were tied, just as they would have been if it were an instrumental line. I haven't checked all my vocal scores, but I'm sure that in most, if not all, cases the quavers are always printed separately for singers, unless there are two quavers for one syllable. Why is this? What are the advantages? Not tying them together looks over fussy to me. Are the singers being encouraged to think of each word separately?

I do mean beamed, in your sense. However, the Oxford Companion to Music, under the article Printing of Music section 2, calls the practice "tied"; i.e. the tails of quavers and such like are joined.

Also, the Bartok example is just repeated notes that get faster then slower. I don't think that all music can be said to be just accels and rits.