Network capacity isn't affected because the amount of data exchanged is the same.
Gas prices isn't affected because gas price is only correlated with the transactions per second (which isn't changed as transaction time isn't changed) and demand and supply for transactions. When lots of people need to make transactions, demand goes up and price goes up. Vice versa. This is the same with other chains too.
Some chains intentionally lower the transaction time to increase throughput to achieve a lower gas fee, but the disadvantage is that only fast peers can verify the transactions fast enough which leads to centralization.
That's good software engineering practice when you have hundreds of billions at stake. Minimize the bug surface.
Furthermore there are other concerns, Ethereum prides itself in running on the most modest of hardware. Many stakers are staking on Raspberry Pi 4.
A blockchain with large transaction speed require very powerful machines. It becomes impossible to sync with less than 1Gbps of bandwidth and storage cost becomes prohibitive.
Doing so before implementing state expiry/rent to reflect the cost of storing data in a blockchain will reduce decentralization as hardware requirements require professionalization of node operators.
Furthermore there is the rollup-centric roadmap for future scalability: https://www.reddit.com/r/ethfinance/comments/r26hhv/rollupce...
Decentralization is good and is challenging to scale. I see eth making the right moves to keep the chain decentralized and accesible to those running nodes at home.
Slowness is not the mechanism, ETH had ~13 second block times before the merge. Bitcoin is slow because the difficulty is targeting longer block times, not because of inherent need for slowness. Difficulty is only required so that adverserial agents (with less than 51% of hashpower) can't rewrite the chain and change transactions.
From a technical perspective, I don't know how a consensus mechanism moving from a system for which slowness/hardness is the mechanism...to instead moving to a system for which slowness need not even be a factor, could possibly not result in faster transaction speeds.