This is not really a bug, more a request to mention in the documentation or document this behavior here:
When reading in a headerless csv-file with 2 columns where the first column contains integers with leading zeros and the second column contains characters, setting keepLeadingZeros = TRUE changes the behavior of fread to switch from recognizing the file as headerless to recognizing a header since the all columns are characters then.

When reading the documentation carefully, this makes perfect sense, but when just checking the keepLeadingZeros part of the documentation and not having the full behavior in mind, this might be confusing (at least it was for me at first).

Adding a sentence to the documentation could resovle this if you don't think this bloats the documentation.