You seem to mainly be interested in checking if PROC IMPORT created a numeric variable from a column whose header name makes it looks like it should be a character variable.
Assuming you asked PROC IMPORT to inspect the whole file before deciding how to read it it can only happen if all of the non-empty values in that column look like valid numbers. For example if you had an ID variable and the ids only include digits. Note that if the column is completely empty then PROC IMPORT will create it as character with a length of $1 since that takes less space in the dataset than a numeric variable.
Once the value has been read into a number there is not a fully reliable way to get back the original text that was read. Your will probably want to re-read the file using the right definition for the variable(s).
PROC IMPORT does not provide any method for that. but it should show the data step it generated. It is pretty simple to copy that code and modify it to define and read the variable correctly and then run the modified code to make a new clean dataset.
If you need to automate this process you might want to skip using PROC IMPORT.
Either write your own code to read the CSV file(s). It is really trivial to read a CSV file. Just point the INFILE statement at it. Define the variables. Attach and NEEDED informats or format. Attach any wanted labels. And then read the variables with an INPUT statement. If you DEFINE the variables in the same order as they appear in the lines of CSV file then you can use a positional variable list in the INPUT statement to avoid having to retype all of the names.
data want;
infile 'myfile.csv' dsd truncover firstobs=2;
length id $20 age 8 sex $1 text $100 ;
input id -- text;
run;
If you really need to GUESS how to read the file then use a tool that has some flexibility to allow you to force the type of variables that are created.
https://github.com/sasutils/macros/blob/master/csv2ds.sas
The OVERRIDES dataset can be used to set metadata used to generate the
code. Only the VARNUM varaible is required and the dataset must be sorted
by VARNUM. The values of INFORMAT and FORMAT must include the period.
Since file is read using list mode input there is no need to include a
width on the INFORMAT. Do not include decimal places on INFORMAT unless
you want INPUT to divide values without periods by that power of 10.
NAME LEN Description
-------- ---- -------------------------------------------------
varnum 8 Column order (req)
name $32 Variable name
length $6 LENGTH as required for LENGTH statement
informat $43 INFORMAT specification used to read from file
format $43 FORMAT to attach to variable in the dataset
label $256 LABEL to attach to variable
