This was a busy week for me, so this solution got a bit less attention. Even still, I think the idea of duplicating a full air/earth reagent is a strong idea, even if its a weak execution. 

My last note was too wordy, so I've moved my compliments here. These metrics are insanely good. I've never considered latency as a metric but it feels like cycles without the pain. On top of that, HC feels less brutal because the solution space is more fixed. I'm super impressed that you pulled off 2 cycles-like metrics without either feeling unapproachable.

I was short on time this week, so this isn't my best work. I think that duplicating the air/earth reagent a ton is a solid method, but this is the wrong way to do it. I'm pretty sure that using the same reagent multiple times would lead to a much faster tape loop.