Dataset only containes empty input spheres error during calibration

[BertinColpronGeoplus]

Occasionally, during calibration, I encounter the following error: "It seems this dataset only containes empty input spheres" ). his issue appears with datasets like NPM3D, S3DIS, and Toronto3D. After some investigation, I believe I've isolated the problem to these lines of code:

# Center point of input region
center_point = pot_points[point_ind, :].reshape(1, -1)

# Add a small noise to center point
if self.set != 'ERF':
    center_point += np.random.normal(scale=self.config.in_radius / 10, size=center_point.shape)

where the chosen potential point is randomly moved around. However, if I understand correctly, it sometimes occur that the point completely drifts away to an area where no points exist at all (sometimes even out of the cloud bounding box!), thus causing the error.

Note that the potential point is modified in place (center_point being a view into the data). I tried to make center_point a copy instead and this effectively avoids the error. But I wonder if this is a valid solution (I am still fairly new to ML and point clouds) and thus would like your opinions.