The vtkMatrix4x4 MultiplyPoint() method most used from Python is single-precision
The method float *vtkMatrix4x4::MultiplyPoint(const float[4])
should use 'double' instead of 'float'. The use of single-precision unexpected to most users of this method, and the similar 'double *vtkTransform::TransformPoint(const double[3])' method already uses double precision.
For VTK 9, I feel that this small break in backwards compatibility is justified.
Note that from C++, most people will not use this signature: they will use vtkMatrix4x4::MultiplyPoint(const float[4], float[4])
or vtkMatrix4x4::MultiplyPoint(const double[4], double[4])
. So C++ users will be mostly unaffected.