Others don't so much eliminate this restriction as they shift the problem to creating so much transaction history that it becomes impractical to run a node that fully verifies said history. In other words, they're significantly less decentralized.
For comparison, "the global average download speed on fixed broadband is 85.73 Mbps as of September 2020"[1] so even if block sizes were triple their historic values it would still be possible for average global users to wait less than a day to complete the initial download.
[0] https://decrypt.co/42427/bitcoin-blockchain-grows-to-300-gig...
[1] https://worldpopulationreview.com/country-rankings/internet-...
A pruned state can be downloaded with no more trust assumptions than that required to trust the software distribution channel where you download the Bitcoin software from to actually serve Bitcoin.
Anyway, every one in the world doesn't need to run a full node. Even a few thousand geographically distributed full nodes is enough to offer resistance to network shut down, and there are more than a few million people that are capable of running a full node of a hypothetical Bitcoin with 1 GB sized blocks, which only requires downloading 1.67 MB/s of transactions to keep up with validation. That's a 1000 expansion of Bitcoin's throughput, letting Bitcoin become the world's most widely used currency.
The entire small block narrative is built on a shaky foundation that ignores the massive benefits of enabling more people to transact directly on the Bitcoin blockchain.
Nonsense. Bitcoin software releases come with widely publicized checksums and signatures from reputable sources.
More importantly, Bitcoin is ideologically opposed to the idea of checkpoints, as they run counter to the Longest Chain Rule that says the true state is whatever valid chain has the most cumulative work. Even if that chain is the result of a very deep reorg that invalidates the last checkpoint.
It's a lesser issue if you can support a large number validating nodes using cheap hardware, they are frequently and randomly assigned to verify different portions of the history, and they always act as secure light clients of the other portions.
In fact, it might even compare favorably to the handful of major mining operations that mostly run Bitcoin, on expensive ASICs.