I'm starting to worry that based preconfirmations lead to too much centralization pressure, creating a small oligopoly of shared sequencers and provers. Is this really better than centralized sequencers (which can also give preconfirmations)? At least the centralization here is per rollup and not global.

An alternative proposal: Accept the centralized sequencer, but make forced inclusions super attractive (same price as normal rollup transactions, just no preconfirmations).

In this case, the sequencer can always give one blob worth of preconfirmations, after which they have to include the "based" forced tx.
Worst case for based tx is to be included after 2 sequencer blobs. (We should also add a max space between sequencer blobs so they can't delay indefinitely)

Here is a possible design:
 - Forced inclusion blobs can be created by anyone (they are "based")
 - They have to be included after the next sequencer blob

20. hacking @ https://www.spire.dev/ 
previously intern Nethermind, epns.io