What you call a fold down is called a downmix in ac3 standards.
The default settings for an ac3 decoder making a downmix of a 5.1 stream to a stereo output is the following (as mentioned here : http://www.dolby.com/us/en/technologies/dolby-metadata.html) :
- Left and Right channels are sent to their respective L and R channels.
- Center channel is sent to both L and R channels with a -3 dB gain.
- Left Surround and Right Surround channels are sent to their
respective L and R channels with a -3 dB gain.
- Usually, LFE channel is ignored.
(Notice there are some differences between a Lt/Rt downmix and a Lo/Ro downmix).
(Also notice that some of the downmix settings are metadata dependant, and you could for example prevent the surround channels to be used in the downmix by setting the appropriate metadata at the encoding stage).
So there's obviously an opportunity for clipping when (for example) summing
L + 0.707 * C + 0.707 * Ls
But the ac3 decoder will usually apply the Dynamic Range Control when downmixing, which is supposed to prevent such clipping. A licensed Dolby hardware decoder will usually apply this control. Software decoders might not implement this control.
Nevertheless, you should also be able to add a second soundtrack while authoring the DVD, you could use a handmade downmix encoded as 2.0 ac3. The end user can then chose which soundtracks he wants to hear trough standards dvd player interface. See your authoring software user manual to achieve this.