If I remember LIBOR correctly, it also had multiple sources of information colluding to influence an "algorithm" that worked based on their claims, whose output in turn was used in other contracts a source of "truth" (which apparently now has caused some issues with changing it, since long-running contracts reference it as such). That seems quite similar.
Right, but as far as I understand it, the problem is that the reported LIBOR and the actual LIBOR could differ, and they could arbitrage that different. I don't think that applies here.
The alleged goal was to make a smart contract that depended on something outside the blockchain by using TLS-N. The LIBOR scandal was an example of a bunch of data sources colluding to lie to manipulate LIBOR. The data sources in this TLS-N scheme could collude to deceive the smart contract.
I don't think so, because these data sources are the exchanges were people trade currencies, and if the numbers sent were significantly different from the rates at people were actually trading, they could be easily caught.
Plus, it's not clear that they have any way of knowing that a certain API call fetching the rates is going to a particular smart contract, so they'd have to lie to everybody in a way that is immediately apparent.
In that case, look at the forex fixing scandal for inspiration. The exchanges reported entirely correct data -- the problem was that banks (allegedly) executed intentionally bogus deals because that only cost a bit of money in comparison to the large amounts of money that they made as a result of manipulating the reported data.
LIBOR was easily manipulated because you could report one level and trade at another. I don't see this applying here, do you?
anyone who compromises the key can forge it
You can mitigate that risk, though, for example by requiring signed rates from many providers.