atomic-revprop: help tracking down ra_serf failures

[ I'm only talking about the ra_serf tests in this email ]

When I build the atomic-revprop branch against serf r1408, I see that:

* diff_tests 24 29 (and many others [1]) fail with HEAD of the branch.
    (IIRC these two failures persist if I switch to serf ^/branches/0.4.x_at_HEAD.)
* diff_tests 24 29 fail with r983269 reverted. (test output attached)
* diff_tests pass with r983269 and r983225 reverted. [2]
* diff_tests 24 29 fail with a working copy @r983225.

Complication:

* All tests passed at the time I committed r983225, using unknown version
  of serf (possibly trunk-as-of-then). I'm sure of this because that
  revision touches nothing except libsvn_ra_serf, and because I believe
  I did run 'make check' before committing it.

Question:

* It seems suspicious to me that a working copy at r983225 passed tests
  when that revision was youngest, but doesn't pass tests now. Does
  anyone see from this what might be causing the failures?

Thanks,

Daniel
(if no one sees why, I'll revert and rewrite r983225 and r983269; but
I hope someone here could save me some work...)

[1] List here: http://svn.apache.org/repos/asf/subversion/branches/atomic-revprop/FAILING-TESTS
    Outputs here: http://ci.apache.org/builders/svn-slik-w2k3-x64-ra/builds/761/steps/Test%20fsfs%2Bserf/logs/faillog

[2] http://svn.apache.org/viewvc?view=revision&revision=983225
    http://svn.apache.org/viewvc?view=revision&revision=983269