I side with "code is law" and would disendorse both EIPs, but I believe innovative fund recovery schemes should be studied.
A sample situation I came up with
If your house got burned down, there's no way to beg the universe to un-burn the house. Most people have home insurance to deal with this problem.
The universe is totally "immutable", as in the valid state transition sense, but blockchains are actually governed by consensus, for example, the DAO fund recovery was authorized by protocol change.
Insurance schemes are proposed in discussions surrounding EIP 867 and EIP 999, and they are application or off-chain solutions.
Solving the problem with better smart contracts
I know immutability is important. I have heard every arguments for it, but I also know that lost funds is sometimes not the victim's fault and "all software have bugs" is something almost every software developer on Earth have accepted. This means, smart contracts themselves need to be amended to fix their security holes. However, echoing the immutability arguments, smart contracts (from 2-party gambling sessions and atomic swaps to complex ERC20 tokens) are contracts and are not meant to be broken, or else they are useless.
Some smart contracts solve this problem with migration schemes, either activated by the contract's owner or through a vote by participants.
On-chain fund recovery scheme
A solution I want to explore is fund recovery through block rewards (also mention in EIP 867 discussions). Every block reward has a % allocated to slowly pay off those who lost their money. The lost fund is paid off in months or in years, so that the right amount of consequences of bad decisions are suffered ($100 fine, not death penalty, for littering). Since hacking and lost funds are socially recognized, on-chain voting to allocate fund recovery seems to be a good idea. But there are details to iron out: What kind of majority is required? What about vote buying (Vitalik thinks every voting system suffers this), voter apathy, privacy of wallet owners, and using it for non-fund recovery situations?). Depending on the voting system, this scheme can also suffer "too big to fail" problem since major dApps are better at campaigning for votes.