Clarify: Transaction selection power-law distribution

[rphmeier]

Quoting from the spec

CONSIDER: Allocating a large CSPRNG subset of validators (maybe 33% + 1) to elect transactions. The subset is ordered with a power-law distribution of transaction allocation. Those allocated greater number of transactions also take a higher priority (and effectively render moot the lower-order validators), meaning that most of the time the first few entrants is enough to get consensus of the transaction set. In the case of a malfunctioning node, the lower-order validators acting in aggregate allow important (e.g. Complaint) transactions to make their way into the block.

Since the transactions need to be included on the relay chain anyway, they have to be subject to some kind of BFT consensus upon inclusion in a relay chain proposal (Tendermint, Aura, PBFT, HoneyBadgerBFT, ...), either leader-based or leader-less.

Is this a way to propagate transactions or to put forward proposals to a leader-less BFT?