So, it turns out that one of the domain controllers had a different certificate chain
(outside of my team's control) which was inexplicably causing the whole thing to
fail.
I would run "ovirt-engine-extensions-tool --log-level=FINEST --log-file=/tmp/aaa.log
aaa login-user --user-name=preader(a)liberty.edu --profile=liberty.edu" and everything
would look fine up until the point that it needed to "doFetchPrincipalRecord",
at which point it would fail to get the principal record for the account. The bind would
succeed, but because "Creating LDAPConnectionPool" would fail on *just one* of
the domain controllers, it for some reason seemed to invalidate all of the entries in that
pool, thereby causing the fetching of principal records to fail even though the bind
succeeded on one of the OK domain controllers.
Is this behavior intended? I really think this should be classified as a bug.
For what it's worth, this was resolved by getting the certificate chain from the
problem DC and then adding it to the Java Keystore with the other certificate chain that
all the other domain controllers use.