Exactly my thoughts. IMO it's analogous to SSL/TLS, which for the longest time was something you had to pay extra for at multiple levels. You had to get a dedicated IP for your service instead of being able to use shared hosting, you had to buy the certificate itself, and you had to have someone renew it and install the new certificate every now and then. Even from the user side there were web sites I remember where SSL access as a client was a premium feature as if it had a meaningful cost per use instead of just the cost to have it available at all.
This was long lamented in the security-focused parts of the internet but no significant movement happened until late 2010 when the Firesheep extension made sniffing unencrypted passwords or session keys accessible to the masses. Suddenly it wasn't just security nerds complaining about things that could easily be brushed off by higher-ups, it was a problem that anyone who could install a Firefox extension (which thanks to ad blocking and the general badness of could see with their own eyes
Almost overnight major services decided that they in fact were fully capable of offering encrypted services to everyone. Suddenly client and server applications, SSL/TLS libraries, etc. that had been ignoring SNI all started supporting it over the next year or two so you didn't need a dedicated IP per hostname anymore. The ISRG was formed and developed Let's Encrypt to solve the problem of having to pay for certs, and while they were at it took the opportunity to develop and enforce the use of ACME with short-lived certificates so automation was effectively mandatory.
In six years we went from there being a "SSL/TLS Tax", which every part of the process would defend as absolutely necessary costs, to it being freely available to all and automatically supported by a lot of major application and service platforms.
I look at the SSO Tax situation as being mostly comparable from the user side of things. The internet as a whole would benefit from wide support for SSO, but the demand to solve the problem is minimal because most of the people making the purchasing decisions don't care about the problem enough to make it a requirement.
There are definitely some valid points that have been made about support costs associated with SSO, especially with regards to supporting more than just a couple of big name identity providers, but I don't think most people would complain if a vendor limited their free/cheap tiers to those couple of big systems those clients are likely already using. If someone needs custom integration they get to pay enterprise prices, but if they just want to sign in with Google that should be accessible to all. The support costs should scale well with a good implementation, hypothetically if Google or Microsoft changes something that breaks your integration you only have to fix it once to solve issues for everyone using it, as opposed to private identity services where each tenant might need something slightly different.