[EarlySNIa] groupby does not apply to objects

[emilleishida]

there was an iminuit error on 19 Feb 2026.

I am unable to reproduce the original error locally (still investigating). However, I discover things that might be related.
This is just a record so I do not loose track of what I thought -- it is not a fix

The one thing different of this data from the previous days is that fluxes can be None.

Because of that, a mask was introduced before feature extraction

however, once this is done, the type of the flux column changes.

One of the things that exist inside the average intraday calculation is a groupby:

df_average = df_average.groupby('MJD').mean(numeric_only=True)

the numeric_only was introduced because of the filter column -- even if this is applied in the same filter. It is probably not elegant but was never a problem.

However, if the flux column is now an object and not a float, the groupy does not work, then returning an empty data frame.

I know the error is not the same I originally started with, but if this is an issue in the broker I can refactor the groupby to remove the numeric_only option