What I do is have HTML messages piped through 'w3m --dump'. 99% of the time this results in a usable representation of the message. The other one percent of the time, it's just another couple of keystrokes to load the content in your web browser.