Surely with today's technology it shouldn't be difficult to build an internally-recording suit that doesn't require a tethered connection. Cell phones prove that GPS technology doesn't require anything too bulky. Why not adapt the idea a bit, and make a suit covered in small sensors that record relative positional data from static transmitters.
I envision establishing a "box" of eight transmitters (that many isn't technically necessary, but might provide more accuracy and error-correction; initial thought is literally a rectangular configuration with a sensor at each point, but the formation need not be that rigid) which is only limited in size and shape by the sensitivity (reception range) of the sensors. Once the transmitters are on, the sensors begin recording relative positions and storing them to distributed flash memory hubs; more hubs = smaller individual storage capacities needed, shorter transmission distances, and possibly smaller footprints. All that needs to be stored would be transmitter ID, sensor ID, relative distance, and a timestamp. The data can be uploaded to a server later, along with inputting the relative positions of the transmitters, and the data points can be calculated and compared to build accurate locations at specific times, and plot out the total motion capture. With a little modification, radio transmission from the hubs could enable real-time uploading to processing server.
Advantages I see to this: transportable (motion capture can be done on location, instead of set stages/rooms), untethered, scalable (use as many sensors as you like). Maybe more that I can't think of right now.
Disadvantages: processing-intensive to convert data points to motion plot.
Unknowns: power consumption of the suit, and how best to provide the power capacity it needs for sustained recording.
Anyone see any major flaws in my admittedly hasty design?