Comment (by Roland Haas):

The test parfile teukolsky.par is the same as WeylScal4's teukolsky.par (2 lines are is changed to output the NPScalar quantities rather than WeylScal4). However WeylScal4 uses test.ccl to reduce he accuracy requirement of the test to 5e-10:

TEST teukolsky
{
  ABSTOL 5e-10
}

and possibly the NPScalar test would pass with that reduced setting. Tolerance was increased in 233494b4 "WeylScal4: Set the absolute tolerance to 5e-10" (9 years ago) by @Peter Diener with a comment:

Since the teukolsky test seem to be very sensitive to roundoff error the
previous tolerance seemed to be too tough. Adding noise at the level of
1e-14 could lead to absolute differences of order 1e-10 and relative
differences at the level of 1e-8.

--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2101/add-lean_public-and-proca-thorns-to-the-et