Part of the reason we see a problem is because thresholds exist. It's called a binding constraint.
There's nothing inherently wrong with having players solve a max f(x,y,z;a) s.t. H>=0 linear optimization problem with a truly binding constraint. We did that on every turn in every Civ up until Civ 4. The mechanic was, coincidentally, called happiness. The system worked, though it forced either hardcore micro or a lot of save and reload.
The hard constraint meant that if you ignored happiness, your empire cratered. You couldn't produce anything at all if you failed the constraint in a city. Civ IV cleverly relaxed the constraint to make it less punitive. What Soren realized is that it doesn't matter how hard the happiness hit is as long as it is
meaningful, so that the player is incentivized to resolve the problem. In this case, Soren just made an unhappy city strictly less efficient than a happy city with one less pop. That's enough.
The problem now is that all production other than food and hammers are additive processes ultimately deposited into the same empire-wide bucket. Right now, none of those buckets are penalized for unhappiness. Since resources are fungible (convertible from one to another), you can substitute resources in your large buckets for the small, city-specific buckets that get nerfed by exceeding the happiness constraint. Since the buckets are additive and unpenalized, conditions must exist where it is desirable to ignore the mechanics entirely.
The most elegant solution to the problem is simply to directly reduce the number of cities that contribute to the empire's productivity as the happiness problem becomes more severe. Under that rule, you will invariably run into a wall where continuing to neglect your happiness problem ceases to pay dividends.
An alternative approach that is likely to be better received is to make extreme happiness problems sufficiently punitive that it will never realistically make sense to expand without limit. It seems that some version of this solution is what most players are advocating.
But it is important to realize that a smooth happiness function will only alter how we manage our empires, not whether or not the game is "fixed" by forcing us to consider happiness. The harshness of the penalty will dictate whether the solution "works". Also, you need to realize that under the current mechanics, any smooth solution guarantees that we will want to run some unhappiness at some times. Either the constraint at the margin is sufficiently harsh that it might as well be a hard constraint, or conditions exist where limited unhappiness is desirable. There is no middle ground.