It doesn't make any sense that you have to leave your job to get a big salary raise. Surely you're a lot more valuable to a company with two years' of experience in a codebase vs. a new company with no experience.
You're not paid your value, you're paid the least amount a corporation can get away with paying you - which is a very different number. (Not that I resent it, it's part of doing business)
When you're being hired both you and the company are taking a bet. The company is betting that you're worth $x+$y/year, where $x is the salary and $y is bit extra to accommodate the risk that you aren't, you're betting that it's worth working for the company for $x-$z/year, where $x is the salary and $z is a bit extra to accommodate the risk that you aren't. I.e. both sides price in risk.
After a year or two, the risk on both sides has largely evaporated. The company now knows that you are (or aren't) worth $x+$y, and you know that it actually is (or isn't) worth working for the company at $x-$z.
There's now a delta between $y and $-z (assuming the initial estimates were accurate) where it "makes sense" for you to stay, the company could pay you more, but why would they? If it wasn't for the fact that shrinking salaries is a negative experience for the employee (more negative than the dollar difference) they could also pay you less. Splitting the difference by not changing the salary at all seems fair, and nicely lines up with the agreement you already have in place.
You're not splitting the difference if market rates change. If pay is going up for marketable skills, the employee is eating the cost, if pay is going down for marketable skills, the employer is eating the cost. If there's inflation, the employee is eating the cost.
General inflation alone needs to be accommodated at bare minimum, otherwise the employee is incurring debt by remaining in place year over year. Most likely by staying somewhere, your skills are also being refined around said employer, so the value of your skills for said employer should be going up.
This ignores all the other factors. Businesses created this sort of employment market with high rates and high turnover because it's what they wanted, it's not the labor force that largely want it, they've just adapted to it.
Change in market place conditions, and change in skills, definitely do create an argument for raising pay. In the tech industry I'd argue that this isn't being accounted for enough (why turn over rates are so high), but if we look at something like "wallmart manager" I think it probably is. This is a good portion of why I started my post with I don't completely disagree.
They should, because they observe the parent's point:
The employee now has the potential to jump ship and re-start this process for a higher $x, meaning that the employer risks losing the institutional knowledge/institutional memory that this employee contributes, and the employer will suffer the cost and risk of hiring someone(s) new.
An employer declining to give a regular pay increase is depending on their employee's loyalty and kindness to remain in employment at lower than their market value, and/or that the average employee has very minimal capacity for risk, because they must constantly stay housed and fed.
That might work for the employer a while, but personally I don't find it a wise long-term strategy.
You can see this phenomenon everywhere and it is as clear as Caribbean water in marriages or relationship: one can be married to the most good-looking, engaging partner around, but doesn't the so-and-so looking person you have met a bar after a few drinks feel like a rush of life? After choosing the rush, some of them have regrets, but it appears that, in the vast majority of cases, life goes on. And it surely goes on for a company, so companies are not very motivated to spend more money for someone who's old news.
My experience in consumer and enterprise tech tells me that most the rationalizations one reads on HN and similar forums, that is the cost-benefit analysis of hiring or retaining, the calculations of how many hours of engineers' time are spent on hiring (or meetings!) instead of working on product, are fun academic discussions as much as doing a cost-benefit analysis of marrying your partner or going on as is, or getting a dog. They all talk about them, but how many do those analyses? How many go by feelings instead, or aspirations?
For example, in my current company (but isn't it the same in all companies?), when somebody leaves, there are the customary "I wish the very best in your future career" and when someone joins there are the customary "It is so exciting to have you here, let's book a 1:1". Opportunistic messages, but they also say something real about how we feel about the old and the new.
I remember a Uber guy who left the company after a few years of tenure. On Twitter, they wrote the usual: "It's been great to be here, I met so many exceptional people (...and all that customary boredom)". After 6 months, another Uber-er wrote on Twitter: "We worked so well on that product, I did not know you left!". It was an enlightening exchange.
The more frequent your turnover, the higher $z becomes.
Applicant A has the precise set of skills Business B needs desperately at that moment. Business B is willing to pay whatever needed to get Applicant A ASAP (because they are growing quickly, or in crisis, or just need that role filled fast). This is normally what causes salary bumps in job switches... because A would not accept an offer if it _wasn't_ substantially above what they make now.