The following worked. I assume it was writing to Excel starting at the index that made up the "first" row.

df3 = df3.iloc[::-3, :]
df3 = df3.iloc[::-1].reset_index(drop=True)