Hmmm, looking at Lojban in a bit more detail it sounds like the consensus is that the conative load of having to construct perfect logical specificity makes it suboptimal as a secondary intermediary language. If people are learning it as a second language it will be very hard to pick up.

This seems like a pretty solid option. I feel like this type of algorithmic language construction could be ripe for a big push forward, both in terms of constructing new languages and benchmarking them for use.