Use in-memory objects to process raster maps using GDAL's Python API

Question

Below are three GDAL supported raster map processing commands, set inside a Bash shell helper function, defined inside a script that'd patch ERA5-Land with ERA5 data (see also: Merge ERA5-Land and ERA5 time series):

function gdalmerge_and_clean() {
    VARIABLE=$1
    YEAR=$2
    BAND=$3
    if [ ! -f "era5_and_land_${VARIABLE}_${YEAR}_band_${BAND}_patched.nc" ] ;then
        if [ ! -f "era5_${VARIABLE}_${YEAR}_band_${BAND}.nc" ] ;then
        gdal_translate \
            -if netCDF \
            -of netCDF \
            -ot Float32 \
            -tr 0.1 0.1 \
            -unscale \
            -b $BAND \
            $ERA5_DATA_PATH/consolidated/${YEAR}/${YEAR}_cds_era5_${VARIABLE}.nc \
            era5_${VARIABLE}_${YEAR}_band_${BAND}.nc
        fi
        if [ ! -f "era5_land_${VARIABLE}_${YEAR}_band_${BAND}.nc" ] ;then
        gdal_translate \
            -if netCDF \
            -of netCDF \
            -ot Float32 \
            -unscale \
            -b $BAND \
            $ERA5LAND_DATA_PATH/cds_era5_land_${VARIABLE}_${YEAR}.nc \
            era5_land_${VARIABLE}_${YEAR}_band_${BAND}.nc
        fi
        gdal_merge.py \
            -of netCDF \
            -ot Float32 \
            -o  era5_and_land_${VARIABLE}_${YEAR}_band_${BAND}_patched.nc \
            era5_${VARIABLE}_${YEAR}_band_${BAND}.nc \
            era5_land_${VARIABLE}_${YEAR}_band_${BAND}.nc
            rm -rf era5_${VARIABLE}_${YEAR}_band_${BAND}.nc era5_land_${VARIABLE}_${YEAR}_band_${BAND}.nc ;
    fi
}

Questions

1. How can I use in-memory (raster map) Python objects and GDAL's Python API to replicate the above commands?

2. While the gdal_translate commands seem easy to implement (after so many Q&As that demonstrate it), is it possible at all to use in-memory object with gdal_merge in a Python script?

3. The source files come in form of netCDF files. Would the use of the netcdf4-python Python interface to the netCDF C library to read-in data speed-up processing?

References

• https://gdal.org/api/python.html
• https://gis.stackexchange.com/a/236754/5256 : gdal_merge
• Is there a reason I should use gdal's Python library instead of just using os.system(gdal command)? : Why use GDAL's Python API
• https://gis.stackexchange.com/a/346716/5256 : gdal_translate, GDAL, Python
• https://gis.stackexchange.com/a/264798/5256 : VSIMEM filesystem (http://www.gdal.org/gdal_virtual_file_systems.html)
• https://gis.stackexchange.com/a/304762/5256 : gdal_translate, gdal.Translate
• https://gis.stackexchange.com/a/304336/5256 : gdal.Translate
• https://gis.stackexchange.com/a/351924/5256 : In-memory file, Create()
• https://gis.stackexchange.com/a/371863/5256 : gdal.Translate, Creation Options
• https://www.programcreek.com/python/example/85545/gdal.GDT_Float32 : gdal.GDT_Float32
• https://gis.stackexchange.com/a/290782/5256 : read/write numpy array, GDAL, Python
• https://gis.stackexchange.com/a/129868/5256 : rasterio