Not the ones I've worked on... Flag as hidden and delete after some number of days is the best practice which I think (and hope) is commonly done?
Practices may vary by organisation and type of removal requests, but the claim that no removals occurs is absolutely false.
For legally-proscribed content, platforms will go to extraordinary lengths to remove not only the backing store but all possible CDN copies of content. This can ammount to millions or billions of items of content (multiple sizes and versions, replicated across multiple geographic locations).
Other instances in which UGC / USC (user generated/supplied content) have been removed are quite public, as with Google's shutdown and data deletion of its Google+ social network.
Perhaps this problem can be solved somehow, but the alternatives have their problems.
I think if you are a business you should choose technology that allows you to be compliant with the law.
Unless the law violates personal liberty, which GDPR does.
I am very curious if you can pinpoint what exactly from it violates personal libery.
It's a cookie cutter rule that restricts contract liberty to create more regimented social interaction.
People should realize that writing a law on a piece of paper doesn't make it real or enforceable
It's a bit like cryptocurrencies and banks. Because it becomes such a big burden for banks (to prove that the money obtained by selling crypto is not related to funding terrorism, crime, etc.), many banks simply refuse service to customers dealing with crypto. Either closing accounts, freezing money, or whatever.
Ultimately, the relevant question is if the root layer of open data systems should allow erasure. There are good arguments to be had on both sides. If you feel that if we don’t have systems that are fundamentally censorproof, the net result will be an eventual all consuming global censorship regime in the limit given enough time, then there is little choice, however.