Block size looks like a simple capacity setting, and every chain that has argued about it has discovered otherwise. The number decides who can afford to verify the network.
Capacity is set by a limit, not by demand
Blocks arrive on a schedule, and each one can hold a bounded amount of data. Throughput is that limit divided by the interval between blocks.
Raising the limit raises capacity in direct proportion. More transactions fit into each block, congestion eases and fees fall while demand stays the same.
This is why the change is proposed whenever a network is busy. The immediate effect is exactly what users are asking for.
The cost lands on whoever verifies
Every full node downloads, validates and stores every block. Doubling block size doubles the bandwidth and the storage that participation requires.
Validation work rises too, since more transactions means more signatures to check and more state to update within the same interval.
A node that cannot keep pace falls behind and stops being able to verify independently, which is the point at which a participant becomes a passive observer.
Verification is where the trust model lives
A chain's guarantee is that anyone can check the rules for themselves rather than taking a provider's word for the balance in an account.
That guarantee weakens as the number of people running nodes falls, because the rules are then enforced by a smaller group whose interests may not match everyone else's.
Hardware improves over time, so the ceiling is not fixed. It rises slowly and predictably, which is why conservative designs raise limits gradually rather than in large steps.
Propagation adds a second constraint
A block must reach the rest of the network quickly. Large blocks take longer to travel and validate, and during that delay other producers are working on the previous block.
That raises the rate of orphaned blocks, wasting work and slightly weakening the chain's security for the same energy or stake expended.
It also favours well-connected producers, since a participant with better bandwidth loses fewer blocks to propagation delay than one on a slower connection.
Why the debate moved to other layers
Because the trade-off is real in both directions, most networks stopped trying to resolve it by choosing a bigger number.
Instead, transactions are executed elsewhere and only compressed data or proofs are written to the base chain, so a fixed block budget serves far more activity.
The constraint has not disappeared. It has been converted into a question about how cheaply activity can be summarised into the space that exists.