Every chain picks a target interval between blocks, and the figures chosen differ by orders of magnitude across networks. The number is not arbitrary; it falls out of how long a block takes to reach everyone.
A block is only useful once it has spread
A block counts for nothing until most participants have received it, verified it and started building on top of it. That takes time, and the time depends on block size and on the quality of the connections between nodes.
Propagation is not instantaneous even on a well connected network. Verification alone requires checking every signature and re-executing every transaction the block contains.
So the interval has to be long enough that a block finishes spreading before the next one is likely to appear. Otherwise the network is routinely working on stale information.
Short intervals produce competing tips
When blocks arrive faster than they propagate, two producers regularly extend the same parent without knowing about each other. Both did honest work, but only one branch survives.
The discarded block is wasted effort. On proof of work chains that means electricity spent for nothing, and on any chain it means throughput that never reached users.
The waste is not evenly distributed. Participants with slower links or worse connectivity lose more often, which quietly pushes rewards toward the best connected operators.
Long intervals push the cost onto users
Stretching the interval reduces that waste, but it lengthens the wait before a transaction lands in any block at all. A user submitting a payment sits idle for the whole gap.
It also makes confirmation depth slower to accumulate. If a recipient wants several blocks of burial before treating a payment as settled, a long interval multiplies the wait.
Chains built around payments and trading therefore lean short, while chains that prioritise conservative settlement lean long. Neither answer is wrong; they are optimising different things.
Block size interacts with the interval
Interval and block size are really one parameter viewed from two angles, because together they set throughput. Doubling the size and doubling the interval leaves capacity unchanged.
Larger blocks take longer to transmit and verify, which reintroduces the propagation problem the longer interval was meant to solve. The two settings constrain each other.
This is why capacity debates rarely resolve by simply raising a limit. The binding constraint is what the slowest participants can keep up with.
Later designs moved the problem elsewhere
Newer consensus designs reduce the penalty for a competing tip by having validators vote on a single chain rather than race to extend it. Wasted blocks become rare.
That shifts the constraint from propagation of blocks to propagation of votes, and the message volume grows with the number of validators.
The trade-off does not disappear, it changes shape. Every design still has to answer how quickly information can cross the network it depends on.