The model behind it relies on a couple of strong assumptions so it does not reflect reality well. It basically assumes all parameters remain unchanged, including the % locked CELO, which is the main influence on Epoch Rewards and the Rewards Multiplier. From what you can see in the data so far, the Rewards Multiplier tends to decrease exponentially, meaning that if nothing changes (again, including % locked CELO) in 5 years it would likely be in the area of 50%.
However, the cLabs Economics team already prepared governance proposals to tweak the parameter settings to provide countermeasures to that downscaling (see the Epoch Rewards Discussion).
I agree with the general sentiment that Validator Rewards should be constant - however redesigning the Epoch Rewards mechanism to allow more sustainable longer-term dynamic should in my opinion be based on more research and careful design choices, which would require some time. From the sentiment here so far bumping target validator rewards to e.g. 85k (would result in ~75k right now) or even a bit more would be a good shorter-term compromise. I am actually in favour of bumping validator rewards a bit more to 95k (would result in ~82k right now). That would compensate a bit for the downscaling so far (75000*(1-0,87)/2 = 4875$ for roughly the last year, assuming a linear downscaling over time for simplicity) and leave a bit more room for further downscaling, so we don’t have to revisit this target again for some time. At the same time validator rewards only make up ~10% of overall epoch rewards, after such an increase probably ~12.5%, so it does not have a meaningful effect on the downscaling itself. That would give some time to research and come up with a longer term proposal - how to allow for a more dynamic locked CELO and validator incentivization without an intervening scaling effect.