Thank you for reading the comments before replying, I am glad you mention that. I do try to cite sources, but my argument about the scalability of the system has not really been covered anywhere that I could find properly. The reason why I can address this issue quite confidently is because I worked in a few p2p projects in the early 2000th and we had all those issues for all the same reasons before, just at a smaller scale — mainly because bandwidth back then was more limited, and the clients created more traffic at the nodes.
I agree with you that the mining belongs into a wider context. You are creating “currency” out of electric energy, producing vast amounts of waste heat. This, in itself, is at the moment not a positive thing, and will not become this way until we deal with both global warming and means of energy production adequately. I have not addressed this before, but I hope I will have time to do so. Working full-time while being a writer somewhat limits the amount of time I can spend on this topic.
Yes, in theory you don’t have to verify every transaction yourself. In order to reconstruct your database, ie the contents of people’s wallets, you do, however, have to download all the existing transactions, or a summary of all these transactions — and trust the summary. The longer period of data you summarize, the more you expose yourself to 51% attacks, since the transactions in the UTXO summary are secured by fewer PoW blocks, thus it becomes less work to rewrite history. This particular solution, and the problem, is mentioned in a paper that is linked in one of the comments — the solution is basically to only summarize transactions deeper than X blocks (Also, the ability to actually verify all transactions in all perpetuity is one of the promises that pops up in the blockchain community since its very beginnings in early 2000s). Once again, this is a trade-off between security/trust and scalability.
Trouble is, however, that even fully summarizing does not save your nodes: every existing full node still has to track every transaction, even a transaction that has not been created by a full node. This means, the traffic at a node scales with the amount of clients, not amount of full nodes. In fact, having fewer full nodes than clients will make the nodes reach the tipping point faster, because the full nodes will also have to serve as “servers” for house-keeping operations, like look-ups on the chain. Here is a more recent article that might explain my scalability reasoning a bit better. Here is another, shorter one that is related to why increasing the amount of nodes on a chain does not make it any cheaper for anyone involved, but might not be very interesting for you because you seem to already understand this.
And of course, fraud is ubiquitous and has happened with every means of payment for as long as there have been humans. But with a credit card, you can at least appeal to, say, VISA, and reverse your transaction. If you have been conned into spending your bitcoin, there is, and can, by the nature of the blockchain, be no recourse — the transaction is final.