The client-selected paths that carry traffic through the Tor relay network are called circuits. Tor clients switch circuits at ten minute intervals for better anonymity. However, latency and bandwidth vary from circuit to circuit, based on the capabilities of the selected relays.
Connectivity and bandwidth are currently (especially with the millions of Mevade guests) too unreliable for such interesting applications as Tahoe-LAFS grids for online storage that's private, anonymous and fault-tolerant. That's especially problematic when mounting grid folders using fuse.
The capability to team multiple circuits would improve both reliability (through failover) and bandwidth. In AlSabah et al. (2013) The Path Less Travelled: Overcoming Tor’s Bottlenecks with Traffic Splitting, the authors "present Conflux, a dynamic traffic-splitting approach that assigns traffic to an overlay path based on its measured latency." They note that "it enhances the load-balancing properties of the network" and "significantly improves the experience of users who watch streaming videos online." They "also show that Conflux introduces only slight tradeoffs between users’ anonymity and performance."
What would it take to implement this in the current Tor network?