Python script with memory for intermediate data: wrong calculations

[IlkaIllers1]

I have a python script that uses the georprocessing tools Copy, Intersect, MultipartToSinglepart, RepairGeometry, CalculateField, GenerateNearTable, Sort, Dissolve, Buffer, AddField, JoinField, and da.UpdateCursor. 

When I write the intermediate data to a Scratch gdb, the model works fine. However, when I write the intermediate data to the /memory, the calculations from the UpdateCursor are not being calculated correctly. The model runs without error messages, but the returned values are wrong. 

Any ideas why this might be happening?

[DanPatterson]

Is that the order of tool use?

Do all the featureclasses reside in the "\memory" workspace (not in_memory) at the time of processing each step (eg Dissolve)

You are have seen the requirements for "\memory" workspaces I presume

Write geoprocessing output to memory—ArcGIS Pro | Documentation

And finally

UpdateCursor—ArcGIS Pro | Documentation

Note:
Using an UpdateCursor on a layer with a joined table is not supported.

Which you seem to have accounted for by using JoinField instead of AddJoin.

Perhaps showing your script would help...although you mention "model towards the end

[IlkaIllers1]

Hello @DanPatterson 

I probably mentioned model because it used to be a model which took a long time to run, which is why I wanted to change it into a python script with the intermediate data stored in the memory. It is rather a long script... probably too long to share all of it. I shared the part where the calculation goes wrong.

# Add and Calculate "Area_ha_in", "Area_perc_in"
arcpy.management.AddField(in_table=_Name_criteria_diss, field_name="Area_ha_in", field_type="DOUBLE")[0]
arcpy.management.AddField(in_table=_Name_criteria_diss , field_name="Area_perc_in", field_type="SHORT")[0]
arcpy.management.AddField(in_table=_Name_criteria_diss, field_name="FID_location_criteria", field_type="TEXT", field_length=1000)[0]
with arcpy.da.UpdateCursor(_Name_criteria_diss, ["Area_ha_in","Area_perc_in","FID_location_criteria","SUM_Shape_Area","Shape_Area_location","FID_common","criteria_ID"]) as cursor:
        for row in cursor:
          z = row[3]/10000
          a = float(str(z).replace(',', '.'))
          b = round(a,2)
          row[0] = b
          u = (row[3]/row[4]) *100
          v = float(str(u).replace(',', '.'))
          w = int(round(v,0))
          row[1] = w
          row[2] = row[5] + row[6]
          cursor.updateRow(row)