Pandas' `to_csv` doesn't honour column data type in one case

Question

Consider the following sequence of operations:

1. Create a data frame with two columns with the following types int64, float64
2. C