The CO2 eventually finds its way from the injection wells to the production wells and comes back to the surface. This is perfectly natural and not a problem; it's allowed for in the design of any EOR scheme. The CO2 is just another part of the well effluent stream, so it's fully contained. You strip it out in the separators, recompress it and reinject it to dissolve out the next lot of oil... and so on. This is called "gas cycling".

Eventually you would have a reservoir full of CO2 (or, more realistically, CO2-rich solvent plus injected water) and some residual trapped oil (probably quite heavy as all the light ends would have vaporized), with smaller and smaller quantities of liquids coming out. When oil production falls below some economic limit, you shut everything down and abandon the wells (you DID provide for abandonment costs in your project plan, didn't you?). The CO2 would remain trapped underground.

It might be good to remind readers here that projects like Weyburn and the Otway basin (Australia) are pilot projects. Yes, CO2 injection has been in use for a long time. Sequestration really is the new part of all this.

The way people are speaking about CO2 injection, capture and storage on this (and other) threads leaves the impression that all the angles have already been worked out (the capture and storage issues). But that's why these are trials with monitoring programs to see if the techniques involve will really work.