It increases the costs of that dataset that cannot be pruned
There's no real reason the dataset cannot be pruned - i've been playing with a DB copy of a blockchain, looking at ways of "removing" the records for accounts with a nil balance (amount out = total amounts in) where date is > 30 days ago
I think you misunderstand. Nobody is saying the blockchain can't be pruned. IT CAN be pruned however the UXTO (set of unspent outputs which can still be inputs for future txs) CAN'T be pruned. That is fine because generally the UXTO is going to grow slower than the blockchain (people tend to spend unspent outputs creating roughly the same number of unspent outputs). There is one exception. That is UNECONOMICAL outputs.
If you have a 0.0000001 output but it would cost 100x as much in fees to spend it would you spend it? Of course not. Kinda like mailing a penny (at a cost of $0.46) to your bank to apply to your mortgage principal. Nobody does that it doesn't make economical sense. So these uneconomically outputs are likely NEVER going to be spent. Each one that is produced won't be spent and thus won't be pruned and will remain in the UXTO forever (or a very long time on average) this is causing the UXTO to bloat and will continue to bloat as there is no reason for anyone to ever spend these outputs (and spending is what allows an output to be pruned).
The UXTO is the critical resources. In order to validate tx quickly the UXTO needs to be in memory. So what happens when the UXTO is 32GB? 64GB? 200GB? Now if those are "valid" outputs likely to be used in future tx well that is just the cost of being a full node. But when 50%, 70%, 95%+ of the outputs are just unspendable garbage it greatly increases the processing requirements of full nodes without any benefit, to anyone.
I don't *need* as a _user_ of bitcoins the whole blockchain, if I could get "balances at point in time" and the journal entries after that.
Of course you don't which is the whole point of pruning the blockchain however you do need to retain a copy of every unspent output otherwise when you receive tx or block containing that output as an input in a new tx you can't validate the tx or block. If the input is coming to you, you can't even know if the tx/block is valid or just some nonsense garbage that an attacker sent to trick you into thinking your got paid.
This unprunable dataset is a subset of the blockchain however tx below the dust thresholding simply bloat this.