It is designed to replace BTC subsidy with fees at some unspecified time, probably in the reasonably-distant future when BTC subsidy is much smaller than it is today. There is zero evidence it was ever designed to replace for example 3600 BTC/day in subsidy.
This can only be a gradual process / not an on-off switch where fees suddenly spike upwards 100x....
Yes but your comment that fees need to rise (as denominated in BTC) is not supported by any evidence, including the chart you linked.
I don't follow... What evidence do you need? That's the design of BTC: Fees have to take over from subsidy. It is a known element of how it works.
There is a glide path in place in reducing subsidy so that fee increases don't go from zero to max overnight when coin distribution is (practically) over. If the designer of the system felt that subsidy must be replaced at an end point, and only then, there would be no glide path in diminishing subsidy. Emission would be flat, at a certain rate, until the last coins are issued.
Do you understand that wallet software and even user education, are part of the system?
If you take the broader view, how about taking an even broader view:
Do you understand that trial and error represent a lesson in user education?
Every user at some point will have a moment where his tx takes a long time because he forgot to pay fees, clicked the "send as zero fee if possible" etc etc. That's the moment when he learned why fees are useful. That's part of his learning.
Now... if that happened once and he thought it won't happen because now he paid 0.0001 / KB, which is 0.01$ for a 250byte tx, then he has something more to learn: It's not only about paying the fee, but how much you pay.
So, taking this even broader view, your client may have learned yet another lesson into how to properly use bitcoin. So the ecosystem improved by improving user education / giving him lessons the hard way.
It's like forgetting your money in your pocket and putting your pants in the washing machine - destroying the paper notes. I'm not going to blame the destructible nature of paper money for that one but I will sure make a mental note to check my pockets in the future, before throwing the pants in the washing machine. Lessons everywhere
Yes it is possible to develop improvements that deal better with fee volatility. Those don't exist (in a well-deployed manner) today, which means there is a mismatch and dysfunction.
Fee calculation is a relatively trivial issue. If wallet devs can't fix it, they better use fixed fees that are more expensive than what should be normally expected but at least ensure first priority for those who desire it.
As an ordinary user who is in no position to fix these things, the system is starting to fail.
EDIT: Also, the wallet didn't use "a very low fee". It used 0.0001/KB which has been a very common fee calculation for at least the past 2 years. It was only slightly below (maybe 20-30% below) what was needed to clear the transaction in a few hours, but remained below for 24 hours.
0.0001/KB = 0.01$ for a 250b tx. It's an extremely low fee and there is no way to spin it around that it isn't.