Probe appears to be missing when interpolating 'bad channels' in recording

Hi,

I am trying to identify and interpolate 'bad' channels in my recording. Whenever I try to save the newly obtained preprocessed interpolated recording, I get the following warning:

"UserWarning: There is no Probe attached to this recording. Creating a dummy one with contact positions
warn("There is no Probe attached to this recording. Creating a dummy one with contact positions")"

There seems to be a probe attached to the recording (The statement recording.has_probe() returns "True"). Whenever I try to save the recording without this interpolation step, I don't get this warning.

Roughly the steps I take:

recording = se.read_openephys(self.oe_path)
recording = recording.set_probe(probe, in_place=False)

bad_channel_ids, channel_labels = spre.detect_bad_channels(recording=recording, method='coherence+psd')
recording = spre.interpolate_bad_channels(recording=recording, bad_channel_ids=bad_channel_ids)

recording.has_probe()
recording_saved = recording.save(folder=session.preprocessed_folder, format='binary')

In a similar issue I read that it is potentially not a problem, since the relevant information (channels locations, groups) are still passed (https://github.com/SpikeInterface/spikeinterface/issues/2895).

Possibly someone knows why I get this specific warning in this case (and whether it is a problem at all).

Much appreciated,

Sietse

Start by reproducing the warning with read_openephys, set_probe, detect_bad_channels, interpolate_bad_channels, and save as shown. Trace how interpolation and binary saving handle probe metadata; done means determining whether the probe is lost or replaced with a dummy and documenting or correcting the behavior.