Consensus of all three users ;)
It is working differently, but not correctly (it sorts Ł as simple L, Š as simple S and so forth). That should not be the case. Is it possible, that the Unicode-sorting scheme is incorrect? It should look like A, B, C, Č, Ć, D, Dź, E, F, G, H, Ch, I, J, K, Ł, L...
Yes, it seems that the ICU library, which MediaWiki uses for collation, doesn't actually support Upper Sorbian in the version used on WMF wikis. Whoops :( I've been able to reproduce the incorrect behavior locally using ICU 52 (the one that comes with Ubuntu 14).
Judging by the dates on http://code.metager.de/source/history/lib/icu/icu/source/data/coll/hsb.txt and http://site.icu-project.org/download/54, it will probably be supported in ICU 54. WMF will surely upgrade to that one at some point, but no idea when that can happen.