One idea to improve proving speeds and increase stack depth is to process (recurse) multiple apps per kernel instead of just one. One main obstacle to this is the rigidity of the current databus implementation which is quite hardcoded based on the assumption that kernels have at most two inputs (one app, one kernel). This is clearly solvable but needs some thought. Are there other obstacles?
One idea to improve proving speeds and increase stack depth is to process (recurse) multiple apps per kernel instead of just one. One main obstacle to this is the rigidity of the current databus implementation which is quite hardcoded based on the assumption that kernels have at most two inputs (one app, one kernel). This is clearly solvable but needs some thought. Are there other obstacles?