I know this is probably the last argument most people want to hear, but is this not a case where more independent implementations would result in less risk?
No. This is nonsense that has been pushed by those actively trying to co-opt the network (or propagated by trolls such as franky). Sure, it would be beneficial to have some competition on e.g. the p2p code but that's about it. More implementations as a side-effects can (and undoubtedly will) lead to even more problems, which will most certainly be harder to solve once multiple node implementations of the network start disagreeing due to whatever reason (be it a bug in this case). <- this is given that you completely ignore that any attempt at a secondary implementation so far has been amateurish at best.Multiple implementations *increase* risk.
^Someone on IRC was pointing out the rather disappointing number of bitcoin sites that were currently managing to expose the bitcoind rpc to the public internet. 
Sadly, yes. Project idea: a open-source, complete web implementation (frontend and whatnot). Back on topic, I think there's two sets of Core users: those who run their node and rarely update it, and the more enthusiastic ones who keep up with upgrades. It might make sense to have a LTS version with more thoroughly tested and vetted consensus critical code (that's proven itself), and a regular version. I think more choice and flexibility could be useful here.
LTS version adoption would make it significantly harder to do a fork bugfix/upgrade whenever it gets needed though.