print "Content-type: text/plain;\r\n\r\n";
[download]

and

    print "Content-type: text/plain; charset=ISO-8859-1\r\n\r\n";
[download]

give different results - no matter which of shell, perl, with or without CGI is involved. For the latter, the browser seems to gather all its input first read in blocks, presumably to apply conversions on the complete text to each. (tested with Firefox only).

Try appending "\0" x 1024 to each line using CGI with print $cgi->header("text/plain").

Another way to confirm that unbuffered IO is turned on with CGI is snarfing the packets with eg wireshark.

...another update: even the trailing ; after text/plain makes a difference. If you are not outputting that, the browser buffers.

_($_=" "x(1<<5)."?\n".q·/)Oo.  G°\        /
                              /\_¯/(q    /
----------------------------  \__(m.====·.(_("always off the crowd"))."·
");sub _{s./.($e="'Itrs `mnsgdq Gdbj O`qkdq")=~y/"-y/#-z/;$e.e && print}