also, i have got a feeling that the gains possible in ledger-entry-processing optimisation are magnitudes higher than in virtual-machine optimisation
THe optimizations have nothing to do with what is stored in the blockchain. ETH's virtual machine is as optimized as any virtual computing environment out there.
The optimizations come in the form of data propagation, node processing speeds (sw optimz, hw utilization, etc). The scaling issue is w/ PoW blockchains, not their particular contents.
The issue really comes down to the sheer amount of data that must be stored on-chain. Moving and processing lots of data is slow, whether it's a ledger or state of a virtual machine.
data propagation is common to both, so it is irrelevant to the comparison. VM is the critical difference; Ethereum's TX processing single threaded & heavy because of it