One of the developers (Alan Fischer) has talked about adding some easy-to-use (but hard-to-implementWhat facilites do you intend provide in regards of network games/simulation ?
(get pos/quat/accel/vel of all objects, time rewind, make simulation deterministic, etc..)

Seems like animation-driven ragdolls are becoming a pretty practical solution. I haven't experimented with this yet, but it should be pretty easy with OPAL. By using a Spring Motor (a simple PD controller that can be used to control position and/or orientation) on each part of the ragdoll, you could simply have the animation sequence drive the desired position/orientation of the Spring Motors.Check this a good addition to your ragdolls : http://www.mmandel.com/gdc
On a related note, my master's thesis research deals with motor learning for physically-simulated creatures, i.e. another way to drive ragdolls (much more difficult, but hopefully more rewarding). Here's a link to some initial results:
http://www.vrac.iastate.edu/~streeter/avh/avh.html
And, if you haven't heard of Natural Motion's Endorphin software, check it out:
www.naturalmotion.com