"nothing relies on it" and "there is not a majority consensus that it is valuable or useful" is not the same assertion. The first is both falsifiabile and demonstrably false. The second may well be true.
Your opinion about the usefulness of DNSSEC is well-documented. That's not the topic of this specific thread: here we are discussing the theoretical case where the DNSSEC KSKs were copied to an unauthorized party, and what damage they may be able to wreak.
As I mentioned, they could use it to forge DNSSEC responses to validating resolvers, now and in the future. It is a fact that some small number of people rely on DNSSEC responses today, so that would be the attack surface.
If DNSSEC usage declines, as it seems you wish it does, then the consequence of such a leak of key material would also decline with that. If it increases, then the consequence of such a leak would also increase, as more people in that hypothetical future would be theoretically vulnerable to forged responses.
If you're a group that steals high value keys for a living, it would be nice to steal the DNSSEC long-term keys when DNSSEC is small and unimportant, because it might be a lot more difficult later if DNSSEC becomes widely adopted to protect critical infrastructure.
Already at least a few dozen people think that these keys are sufficiently important to spend thousands of dollars and dozens of valuable witness-hours traveling to these key ceremonies and storing these safes and whatnot. Perhaps they are entirely misguided and should abandon the entire endeavor.
I don't think the "DNS TLDs are simply a proxy for governments" is enough of a disqualifying argument when several governments or even large corporations have implicitly trusted CAs in widespread use.
I think it's useful to have another choice of validated data to cross-check with. Perhaps you are worried that it would de facto replace the existing corporate-managed PKI with a government-managed one entirely?
I also think that what another commenter mentioned, that you have a choice in governments (and could theoretically publish your signed certificate data in multiple zones, controlled by multiple different governments) is an extremely beneficial set of possibilities. Multiple-TLD-publishing is certainly a better system than the attempts at public-key pinning and observatories/transparency we've seen as attempts to solve the "anyone can issue for any domain" problem as it stands.
...unless, of course, the root zone KSKs get stolen. Then all the TLD zones are unsafe. :)