Adding more stuff to do on the mission increases the danger during the mission, and increases the chance something will go wrong. Anything that goes wrong, even if it's a redundant system to prevent another system from failing completely, can create problems for the rest of the craft depending on how it fails.

So, it would be a question of "where do we stop" in terms of adding more crap to the hubble (which we did fairly routinely for a bit) versus balancing the risk added.

I agree with your comment as a general caution but it seems to me unlikely a redundant system would be any more difficult to install than a one off and balanced against the possibility of eliminating a much more chancy need for multiple trips it seems to me the redundant system would be a worthwhile investment.

That is a good question considering how much redundancy NASA puts in their other projects. I am not sure but I have to imagine the new set of batteries will last 16+ years. Putting numbers to paper, and looking in NASA crystal ball, you might find two things. First, sixteen years down the road NASA may believe they could launch a more productive telescope for less than it would cost to maintain the Hubble, hence no need to have another set installed now. Then there is the issue of those batteries... I have to hope that in 16 years they have a cheaper, lighter and more robust battery alternative then what you see today. But in the end the agency probably just doesn't feel the added cost of having a back up equals what the Hubble is actually worth.

There is another space telescope in the works but it covers a different range of the electro-magnetic spectrum. It is an infra-red instrument and will not replace but will complement the Hubble visible spectrum scope. These instrument take decades to design and construct and to permit the Hubble to be destroyed is to lose an instrument that has an irreplaceable value in the current tight budget economy. In addition, instead of a totally new instrument, the Hubble has repeatedly been updated to increase its capability so to lose it would be a major tragedy.

The flight package has specific dimensions and weight distribution. So whatever needs replacement needs to fit within that volume. If you've ever seen a flight package, there isn't much, if any, extra room. The flight package will be shielded and have thermal conductors to route heating and cooling to the components. Also, by changing the weight distributions, the software would need to be adjusted to allow for automatic tracking of distant targets. This may be simple, or it may be an expensive effort in non-linear equations.
I suppose if there was a little room for traditional 3-axis controls using 3 gyros, a redundant set of 2 gyro controls might be incorporated. I know a post-doc that spent two years trying to reduce gyroscopic controls from 3 gyros down to 2 using non-linear equations. I don't know what was the outcome.

Since the 2 gyro version uses rotary acceleration as one of the components for 3-axis control, I would have to assume it would use more power as well. This would mean the batteries might not last as long.

I'm kind of surprised they didn't mount a thermionic reactor on the Hubble. Topaz

Adding redundant systems to anything is akin to the problem facing small aircraft pilots: are two engines safer than one? Experience has shown that twins have a higher rate of fatalities than singles, not exactly what you would expect.

Generally it boils down to pilot error, but doubling the number of engines doubles the possibility of engine failure as well.

Let face it: the Hubble started with its share of "pilot error".

I sincerely doubt the analogy of single and twin engined aircraft can be generalized to all possible redundancies.

That's absolutely correct - all redundancies are not equal.

However, it's impossible to say that "redundant" means, for sure, twice as safe. If you are adding two engines, and the engine has X change for exploding in a nuclear fireball, annihilating the planet, putting two engines on reduces the chance of plane falling out of the sky, but it also increases the chance that there will be a nuclear holocaust.

While that example is clearly exaggerated, the Hubble's systems are not in isolation. If a spare gyro short circuits, and comes online after being bombarded by radiation, that is not good. Nobody could point the Hubble at anything. If a spare battery explodes in the same conditions, it's the same problem - the rest of the system is compromised. These are also bizarre, and unlikely situations. But NASA has to account for weird crap like gyros shattering, and other nonsense.

"N+1" thinking works for computer systems, but the safety advantages do not translate over one for one to the real world.