Right.... if they have 25, the hard part is done. Now, it's just set up the machine and press go.
Well if the initial production run were done on a MOSIS process (partial wafer with other users), they would need to tape out again for a full wafer, unless they wanted to spin the same wafer over again. The tape out is what costs the most, but having a proven-working design certainly helps things along when you want to graduate up to full wafers.
I assume that the company doing the multi client wafer would have a schedule and would be doing another one the next month or whatever period they run on. Assuming their test worked, they could buy multiple slots on the next month's shared wafer and have 25*n chips, n being any number that suits their fancy. And in that month while they waited for that production, they could put together all of those rigs, so by the time the chips themselves are ready, they can be dropped right into the rigs and fired up immediately.