If I zoom your chart out, I could make all the same arguments about 2018. It looks like just another cyclical metric that follows the halvings.

In the year after the 2018 peak, fees approached zero, but in the year after the 2021 peak, they stayed higher. As long as that keeps up and they don't actually go to zero, and people are competing for block space, miners will keep lining up, and the network will be fine.

Halvings matter less as time goes on. I didn't shed any tears for miner incentives when the reward went down by 25, and I won't shed any tears when it goes down by a mere 3.125. Mining was never meant to be profitable beyond BTC's bootstrapping phase. In fact the protocol actively makes difficulty adjustments to keep profitability near zero. Let their profits go down. As long as people are competing for block space, and transaction fees keep coming in, miners will continue competing for their fractions of a percent, and the network will be fine.

I don't understand the relation you're alluding to between halvings and transaction fees. As far as I'm concerned, they're entirely unrelated. It's true that both big spikes in transaction fees happened to occur four years apart and in the vicinity of halvings, but that could just be coincidental -- if it's not, I don't understand the mechanism that would cause it.

In any case, neither spike was particularly long-lasting, and even if they were sustained at their peak, neither would make up for the mining incentive lost in the next halving.

I only brought up halvings because it explained the chart. But let's ignore halvings and talk about mining incentive then. Fees are the less important half of the story. The main driver is difficulty adjustments. If scarce fees come in, difficulty goes down so profitability increases. If plentiful fees come in, difficulty will go up so profitability decreases. Fees are actually damped by the adjustments so the absolute amount of fees doesn't really matter. So long as fees are above 0.00000000, the network will be fine. I predict that there will always be some baseline level of competition keeping fees above 0.00000000, but if that is wrong, then things would likely fall apart one day.

The problem is, that when it become inprofitable -- miners would shut down their ASICs and wait for better time -- causing difficulty to drop.

And at this point somebody with huge computing power disabled could enable theirs Rig and quickly generating block to perform double-spend, and create soft fork.

The security of the network depends on the difficulty. It’s true that in the absence of bad actors the network will run as long as fees are nonzero, but the whole point of bitcoin is that it’s supposed to be resilient to attacks.

If usage is low - fees will be low, and the cost of an attack will be low, but the expected value gained from an attack will also be low.

If usage is high - fees will be high, and the cost of an attack will be high.

Security scales with usage, as does the incentive to attack. As long as usage trends in the same direction as value, I don't see the problem.

Exactly. Let supply and demand meet and set the price per transaction acceptable to the miners.