You can find the full text of the NTSB investigation here and according to that report (bolded for emphasis),
Examinations of the left propeller components indicated a propeller
blade angle of about 3 degrees at impact. This position was based upon
the position of the pitchlock acme screw. The left PCU ballscrew
position indicated that the PCU had commanded a blade angle of 79.2
degrees. The discrepancy between the ballscrew position and the
position of the pitchlock acme screw is a strong indication that a
disconnect between these two components occurred prior to impact and
that the left propeller had achieved an uncommanded blade angle below
the normal flight range.
And later in the report
Using measurements and the inspection procedures for the quill and
transfer tube of the Hamilton Standard Alert Service Bulletin, it was
determined that the left PCU quill spline was worn to the extent that
its gear teeth did not engage the transfer tube spline. In addition,
the test cell and flight tests showed that the propeller blade angle
could not be controlled by the PCU with a disengaged transfer tube. In
the test cell, the blade angle moved toward high pitch; however, the
propeller was operating at zero airspeed and did not experience normal
flight loads. In contrast, the flight tests showed that the blade
angle would move toward low pitch with a disengaged transfer tube. The
blade characteristics indicate that centrifugal and aerodynamic
twisting moments tend to move the blades toward low pitch.
Board believes that the worn quill on the left engine PCU became
disengaged from the transfer tube prior to the loss of control of the
airplane during the approach to Brunswick. Moreover, the propeller
blades moved to a low angle, resulting in an asymmetric lift and drag
condition that exceeded the capability of the pilots to counteract
with the airplane controls available.
In simple terms the unit that controls the propellor pitch experienced a failure from worn out components that lead to a case where the propellor did not completely feather. The specific part seems to be the PCU quill (to answer your question directly).
Even later in the report they discuss why the failure was not really covered by the testing originally. The testing was not necessarily incorrect as much as it may have been incomplete and not foreseen this issue with the quill.
The investigation found that wear of the quill was not considered
during the certification of the propeller system because of the very
light torque loading on the quill during flight. Service history of
the PCU quill prior to the introduction of the titanium-nitrided
transfer tube indicted that quill spline wear was not a problem.
Additionally, the manufacturer provided an analysis during
certification indicating that even in the event of a failure, the
propeller would either drift into the feathered position or maintain
the blade angle present when the failure occurred. However, the
accident involving flight 2311 and the subsequent investigation
have determined that these assumptions, though originally supported by
numerous engineering evaluations and manufacturing experience, are
invalid and that there are single failure modes that could result in
an uncommanded propeller blade angles below flight idle.