So far, your options are limited to spawning another process (via fork or some Win32::Process-enabled system call) or using some sort of keepalive technique to keep sending data to the client.

The former is better than the latter, but here's another idea. Have your CGI program only collect the data, and write it to a special directory somewhere (or a file, socket, pipe, or whatever). Have a second process (not connected to the web server in any way) do the actual processing. Your CGI can exit and close the browser connection right after it finishes collecting and storing the information, while the other process can do what it needs to do after the user has gone on to something else.