The failure seems to be triggered by some other changes.
First in b134 the font configuration file for this RH/OEL release was
removed, so that it falls back to the synthesised font properties file.
Although I haven't yet figured out exactly which fonts are now being picked up
instead that plays a role.
But even using a particular font from OEL 5.6 that causes a failure
of this test when used anywhere (eg Solaris), that failure in turn seems
to start after b61 in b62, 63 or b64 (we've lost b62, b63 forever it seems) but as
6850113 : Bidi class needs to be updated to support Unicode 5.1
was in b62 I'm guessing it started then.
So b134 changes unmasked a bug that actually goes back to b62.