I think the logic is that 2.9.4 is so horribly broken that you can't even $FORCE it

No, I think it's only breakage is the schema regression that was causing some tests in (t/40reader.t) to fail.
From https://rt.cpan.org/Public/Bug/Display.html?id=114638:

Test Summary Report
-------------------
t/40reader.t                                     (Wstat: 512 Tests: 89
+ Failed: 0)
  Non-zero exit status: 2
  Parse errors: Bad plan.  You planned 100 tests but ran 89.
Files=71, Tests=2552,  2 wallclock secs ( 0.19 usr  0.04 sys +  1.95 c
+usr  0.20 csys =  2.38 CPU)
Result: FAIL
Failed 1/71 test programs. 0/2552 subtests failed.
make: *** [test_dynamic] Error 255
[download]

As Anonymous Monk pointed out, this problem has now been worked around (as of version 2.016).
Hence cpan-testers will give it a "PASS", so allowing them to test it (by not doing the 'exit 0') won't result in any FAIL reports.

Version 2.9.4 is still on the blacklist solely because of that regression - and the workaround was simply to skip the failing tests.

Cheers,
Rob