which I assert the control of must be centralized in order to not diverge.

BTW, do you know that the money is impossible in a completely decentralized world? Because a monetary system can exist only within boundaries of an economic cluster which implies a non-zero level of centralization around the core.

Can you rephrase that statement? I'm having trouble parsing it; it sounds like you are saying opposite things one after the other.

I think my point still applies. What is his incentive to include more than the single partition he is validating in his blocks?

By building on top of another partition's block, he increases the risk that his block will be orphaned because he has cannot tell if it contains a double spend. For him it is better not to build on the best block at all, but to maintain his own partitioned chain of blocks.

They are not independent when you combine them like that with the LCR rule. You must see that? Every block that partition A builds on top of partition B's block is essentially a merger of partitions A and B.

As I understand it, gas is paid within the script activating transaction.

You're putting words into my mouth there - I agree that there is a disincentive to merge partitions yet partition merger is essential for consensus; that is the basic problem facing all trustless p2p currencies and is particularly acute when there is no mining incentive.

Ethereum is a system of redundantly executed code, a general purpose CPU mirroring the functionality of the worlds most inefficient botnet without parallelization.  Unless blockchain scaling is either huge or infinite, the validation and redundancy of simple transactional data, which Bitcoin attempted to solve in a distributed manner, is far more useful in a cost benefit analysis rather than open ended computational redundancy.  In other words, Bitcoin is an ASIC for a specific purpose, and struggles to fulfill that role even with a streamlined design.  Ethereum, on the other hand, has introduced an objectively less efficient design to try and conquer even more problems.

Such a design would obviously be dead on arrival out of the gate barring any other factors being introduced.  This is where partitioning comes in.  A partitioned network allows load to be split in order to scale further.  Here comes the paradox.  If the entire purpose of Ethereum is a system of redundantly executed code, how does partitioning even fit in the picture in the first place?  I guess you can say, ok, we're arbitrarily defining how much redundancy is enough, but how do you actually determine that?

As you can see, Ethereum is effectively useless without partitioning because it relies on scalability far more than Bitcoin does, while Bitcoin is struggling to achieve any scalability itself.  When you see Mike Hearn talking about using Ethereum as an OTC derivatives settlement, that functionality will be non-existent without partitioning.  The load carrying capacity would be far too small and fees far too large to do at scale.  It will be a case of centralized services vastly outperforming decentralized ones.  Who cares if it's trustless if it costs 10,000 times more?

Transactions can be bundled in Bitcoin without much of a side effect in order to scale, while bundling computational redundancy in Ethereum would defeat the purpose of it existing at all.  This is in addition to the fact that nobody has successfully deployed a partitioned cryptocurrency in the first place.  Fuserleer has been working on one for years and has been unable to release something.  IOTA is about to release.  I haven't dug through it enough myself, but Anonymint claims it's doomed because it will rely on centralization for convergence.

In summary, through bundling or even just plain on-chain transactions alone, Bitcoin can just barely reach functionality of an internationally used currency, at least justifying it's market cap.  Ethereum operating as a "world computer" is an obvious stretch of the imagination.  Since they issued an IPO money grab for their coin supply, I think most investors will already be considering the coin a failure if it has to dilute investors to infinity by leaking mined coins forever.  No actual secure PoS network has even been invented so far, so the coin needs not only one, but two cryptocurrency firsts in order to be considered viable - a valid PoS network and a partitioned network.  In order to even think about touching Ethereum, you're basically required to ask yourself, is Vitalik an order of magnitude smarter than Satoshi?