I am confused here.

Is the reason you are distributing binaries that have the different
definition of FAIL() that you expect a subsequent update to the 2.3
image to have the required "auto-disable-on-primitive-fail" very soon?

At this point, I'd favor your distributing interim VM's with the
work-around definition of FAIL so that people with UNIX hosts can
actually run Squeak 2.3.  (The cost is fetching another 150K image.)

Am I failing to understand something here?  Is it intentional that all
UNIX versions of Squeak should randomly hang when the user trips over an
object that happens to play a sound?  Am I just using Squeak incorrectly?

Two out of four systems that I'm running Squeak on are Linux, so I'd be
happy to run with a sound-enabled beta.  If the only way I'm gonna get a
usable 2.3 squeak is to compile it myself, point me at sources that have
the Linux sound patches too.

-wdc