I'm sure I someone will come along and call me paranoid for having such views. Considering current events, I don't think a little paranoia is a bad thing. I prefer to look at it as being prepared. Erring on the side of decentralization is the prudent choice in my opinion.
I don't consider your 'paranoia' unwarranted. However, I think keeping the transaction processing capability down works against your stated goals.
It would seem to me that your stated goal for 'decentralization' is an increase in the number of non-mining nodes. I think we can likely agree that the only population that would have any interest whatsoever in being a non-mining node would be Bitcoiners. I reason from this that increasing the number of Bitcoiners would add the the population of people that may self-select in becoming non-mining nodes. Indeed, it would be my expectation that increasing the number of Bitcoiners will -- all else being equal -- lead to more non-mining nodes.
It seems axiomatic to me that making Bitcoin applicable to more use cases, or more usable for the use cases it has, or downright cheaper to use, will lead to more usage, and more Bitcoiners. And that increasing transaction capability and decreasing transaction cost both serve these aims.
Long story short, I believe making Bitcoin more capable -- even at some real increment of per-node operational cost -- will lead to an increase in the number of non-mining nodes. I realize others may differ. And creating a controlled experiment is likely out of the question.
Of course, I really don't think non-mining nodes add any value to the network. Only to their operators. Again, I realize opinions on this matter differ as well.