There seems to be something wrong when resampling the data, but the problem also exists on main. When loading the file sub-13_ses-S001_task-HCT_run-001_eeg.xdf, the signal scaling (at least for the EEG channels) changes as well as the location of the gap:

Using the default of 52 Hz:

Using 256 Hz:

It looks like the first one shows the correct signals. I'm not sure what the problem is, because the streams are always resampled (this is necessary when loading 2 or more streams). @DominiqueMakowski maybe you have time to look?

I've tested with NK and I it works "as expected"

import neurokit2 as nk

df, sr = nk.read_xdf("sub-13_ses-S001_task-HCT_run-001_eeg.xdf", upsample=0.25)
df[["TP9", "TP10", "LUX", "ECGBIT1"]].iloc[0:60000].plot(subplots=True)

df, sr = nk.read_xdf("sub-13_ses-S001_task-HCT_run-001_eeg.xdf", upsample=0.5)
df[["TP9", "TP10", "LUX", "ECGBIT1"]].iloc[0:120000].plot(subplots=True)

The scaling issue is quite odd, but probably related to some internal handling

Wait, but here you're using your own (=NK) implementation, right? You did not compare it to what MNELAB is doing in this PR?

well no but to be honest I don't really have a good grasp on what mnelab does for xdf 😕