In fact, the slashing amounts are not meaningfully different under Oxford.
When a new cycle begins, the protocol computes a value for the amont at stake. Then:
- double bake incurs a penalty of 10% of this value per double bake event (before oxford, it was 640 tez).
- double attest (also known as double endorsement) incurs a penalty of 50% of this value - this did not change in oxford
We should distinguish between double-bake and double-attest. Double bake is much rarer than double attest, as you need to have baking rights at the exact time you are equivocating. But only two double-attest denunciations are enough to completely wipe your stake.
Most accidental equivocation is due to the operator having two identical baking setups connected to the network. In this case, the most likely scenario is that you would double-attest before you double bake, wiping 50% of your stake in the process (actually, probably 100% unless you switch your baker off really fast). This penalty has been in place since Ithaca and tenderbake. Notably, no slashing has happened since Ithaca.
The switch from 640 tez to 10% is technical in nature: due to the introduction of Adaptive Issuance toggle and the new staking mechanism, it was necessary to have a fraction rather than a fixed amount. 640 tez corresponds to 10% of the stake of a small baker. For a large baker, it is indeed more severe. But again: you are most likely to double-attest before you double bake, so it largely does not matter.
Moreover, what's the purpose of extending slashing penalties to stakers?
Stakers earn more than delegators, because they take more risk: they risk being slashed if their baker equivocates. In case of double bake, both baker and every staker gets slashed 10%. In case of double attest, they all get slashed 50% of this value.