Upstream says:

I think you can ignore this. It's almost certainly to do with minor
difference in locales between OSX and Linux. You are using the same
locale as I am (en_GB.UTF-8) in the OS and this usually means that the
collation tables for the locale are different between OSes. Since the
default is to use the Unicode Collation algorithm anyway, this hardly
matters. I only use OS locales in the tests to speed them up a bit.

https://github.com/plk/biber/issues/72#issuecomment-114232856

---

diff -u -r biblatex-biber-2.1-orig/t/sort-order.t biblatex-biber-2.1/t/sort-order.t
--- biblatex-biber-2.1-orig/t/sort-order.t
+++ biblatex-biber-2.1/t/sort-order.t
@@ -4,7 +4,7 @@
 use utf8;
 no warnings 'utf8';
 
-use Test::More tests => 15;
+use Test::More tests => 14;
 
 use Biber;
 use Biber::Output::bbl;
@@ -482,26 +482,6 @@
 is_deeply([$main->get_keys], ['L9','L6','L7','L8','L5','L4','L3','L2','L1B','L1A','L1'], 'nty with descending n');
 
 
-# testing case sensitive with fastsort
-# In alphabetic, all uppercase comes before lower so the
-# "sortcase => 1" on location means that "edinburgh" sorts at the end after "London"
-# Take this out of the location sorting spec and it fails as it should
-$S = { spec => [
-                                                    [
-                                                     {sortcase => 1},
-                                                     {'location'     => {}}
-                                                    ]
-                                                   ]};
-
-$main->set_sortscheme($S);
-
-$biber->set_output_obj(Biber::Output::bbl->new());
-# Have to set locale to something which understands lexical/case differences for this test
-# otherwise testing on Windows doesn't work ...
-Biber::Config->setoption('sortlocale', 'en_GB.UTF-8');
-$biber->prepare;
-$section = $biber->sections->get_section(0);
-is_deeply([$main->get_keys], ['L1B','L1','L1A','L2','L3','L4','L5','L7','L8','L9','L6'], 'location - sortcase=1');
 
 # Test nosort option
 $S = { spec => [