You would have two things going on: a long-running process that does not communicate with the browser, and an auto-updating process spawned by the long-running one.

The long-running process would write a status file periodically, and the fast-updating page would check that status file and relay the information to the browser, reloading regularly. When the status file indicate the process was done, then the relaying process would send the final results in a page that doesn't auto-reload.