• poseToVector() was added in [wpimath] Add core State-space classes #2614 and never used
• poseTo3dVector() was added in Add pose estimators #2867 to turn a Pose2d into a Matrix<N3, N1> for the KalmanFilterLatencyCompensator global measurement vector (the pose estimator class used a Kalman filter class at the time)
• poseTo4dVector() was added in Add pose estimators #2867 and never used

poseTo3dVector() doesn't save much typing

public static Matrix<N3, N1> poseTo3dVector(Pose2d pose) {
  return VecBuilder.fill(
    pose.getTranslation().getX(),
    pose.getTranslation().getY(),
    pose.getRotation().getRadians()
  );
}

for a class that already has a lot of moving parts:

public void addVisionMeasurement(Pose2d visionRobotPoseMeters, double timestampSeconds) {
  m_latencyCompensator.applyPastGlobalMeasurement(
          Nat.N3(),
          m_observer, m_nominalDt,
          StateSpaceUtil.poseTo3dVector(visionRobotPoseMeters),
          m_visionCorrect,
          timestampSeconds
  );
}

I don't see a use for these functions outside this very specific instance. They aren't even useful for affine transformations directly, as you pointed out.

Unfortunately, we can't merge it until we drop GCC 10, cuz we hit a Raspberry Pi compiler ICE in unrelated code.
https://github.com/wpilibsuite/allwpilib/actions/runs/12015767540/job/33494553453?pr=7430