At the moment I think the only way you could join the conversation is if someone on the far end conferences you in. You would need a PBX or a sip account with at least 2 channels. If you are not at the far end, how would you ever know when to barge in?
Years before there ever was an Obi, I had a serial voice/fax modem listening in to catch caller-ID and a daemon polling it for updates. I would thus see when a call came in, but also I was able to add to it to instant message me. The technology for doing this should be even easier now.
I think we might have a modern and more typical scenario where someone at home answers a call and really another family member should join in on conference. Contact the family member by a cell to cell family plan call to tell them to barge in, then let them connect in with full fidelity of a SIP phone call.
Bridging the call with 3-way calling to the cell phone might work with some fidelity drop as well as cell phone charges. For the case where I have data bandwidth available and I can barge in I think its better all around and a worthwhile feature to consider having.
What is the limiting factor in the current system?
Since the auto attendant allows a trusted remote caller to place a call over the PSTN line, clearly the Obi hardware has the capability to join the virtual and physical ports. Perhaps there is a code to "blind dial" (not wait for dial tone) a single digit that would provide this feature?