Fees on a public blockchain are not a price list. They are bids in a continuous auction for a fixed quantity of space, which is why the cost of an identical transaction varies enormously over time.
Block space is the scarce resource
A blockchain produces blocks at roughly fixed intervals, and each block holds a limited amount of data. That limit is a deliberate design constraint rather than a technical accident.
The constraint exists so that ordinary participants can verify the chain without specialised infrastructure, which is what keeps validation distributed.
The consequence is that throughput has a ceiling, and demand above that ceiling cannot be met by producing more capacity on demand.
Fees are bids, not prices
Pending transactions wait in a queue, and those assembling blocks generally select the ones offering the most compensation per unit of space.
A transaction offering too little is not rejected. It waits, potentially for a long time, until demand falls enough for it to become attractive.
Because the bid buys space rather than value transferred, moving a large amount can cost the same as moving a small one, while a complex transaction costs more than a simple one regardless of value.
Why costs spike rather than drift
Demand for block space arrives in bursts driven by market activity, launches and periods of volatility, while supply is fixed by the protocol.
When demand exceeds capacity, the queue grows and the bid required to be included rises quickly, because users competing for the same slots outbid each other.
The spike ends when demand falls rather than when supply responds, which is why elevated fee periods resolve abruptly rather than gradually.
How fee mechanisms have been redesigned
Some networks moved from pure auctions to mechanisms with an algorithmic base fee that adjusts according to how full recent blocks were, plus an optional priority payment.
This makes the required fee more predictable in the short term, since users can observe the base rate rather than guessing at competing bids.
It does not create capacity. Congestion still raises costs, and the mechanism changes how the increase is communicated rather than whether it happens.
Why scaling moves activity off the main chain
Secondary layers batch many transactions together and record a compressed summary on the base chain, so the scarce space is shared across far more activity.
The economics change accordingly, with users paying a small share of one base-layer transaction rather than competing for a slot individually.
The trade-off is added complexity and a different set of assumptions about how funds are secured, which varies considerably between designs.