May need to consider satellite platform and date when doing group by solar day.
This will be needed for geomedian - if pulling in landsat 7 and 8 on jan 01, one image is lost (the last).

Implement arcpy scripting module cancel monitoring

New bug with gdal.Open, no longer opens WCS queries from within Download class. Crashes. Tried various things including /vsicurl/ and WCS: before url, doesn't work. Former works in Pro 3.2 notebook.

Added temp (?) fix by downloading tif as a file first, then opening within GDAL. May end up being safer - less stored in memory. Current method throttles cpu more, may want to look at requests(stream=True)

S2Cloudless mask excludes too many scenes even when using for e.g. 50% cloud. This is likely due to the current lack of consideration for the probability mask layer - need to implement.

Need to check this. I think the oa fmask isn't working.

Need to implement basic (for now) outlier removal. Z-score?

Boolean not being read?

Will be needed for geomedian.

Not sure how to handle this, as we lose satellite info on first netcdf write. Might need a dim/coord holding sat info for each datetime slice.

Alternatively, retain time in datetime dim and just output that in filename too.

Can iterate over vars and do work (e.g., interpolate) to allow progress bar to be implemented.
See outlier for example.
Also, set -999 to nan in loop instead of all at once at start - could reduce load time and memory size.

Removing SLC-off data should not be downloaded by mask function. Look at this.

Check if this is working - seems to speed through now. Usually slow.

With dask we need rechunking etc

Likely a data type issue. Look into it.