I'd write a cron job that periodically checks the httpd.pid file, tests to see if it's valid, restarting the server and sending a mail if not. Monitoring it from another location with LWP is a good idea, but assumes that the network links between the second box and the first are reliable. (That may or may not be an assumption you're comfortable making.)

O'Reilly's Perl for System Administrators suggests not sending mail directly from a monitoring script, if you have more than one. Rather, aggregate all of the responses and just send one e-mail with all of the emergency messages.