This sounds like the wrong solution. If the issue was avoiding some techs as a diplo civ, those units need to be made better instead of making diplo units artificially worse just because you advanced an era, even though nothing else has changed. Trade units scale in cost because the value of trading increases over time, a caravan in Ancient era is just a couple of gold while Information era it could be 40+. But diplo units are actual units unlocked on the tree.
They should be good enough that you'd want to tech and build them, and then scale the production to their increased value, just like we do with all military units without issue.
The problem (from last year):
1) You have more and more production and gold as time advance.
2) Which mean a single diplomatic units is "cheaper and cheaper" as time advance (100
is not worth the same at turn 50 or turn 500).
3) However, we do
not want inflation on influence. We would like effects like "+1 influence per turn" with a city state should remain relevant in the late game. We want the "30 influence for friend" to not be something you have for almost nothing, even in late game.
There was two solutions:
1) Make everything scale. The threshold for influence would scale with era. Most effect that give influence would scale with era. Except diplomatic units: influence from diplomatic units would scale only by upgrading them.
=> This solution would have required additional coding, in order to modify all the effects. Moreover, having influences values over 1000 the end game would feel quite stupid.
2) Make diplomatic units worse and worse as time advance, in term of production vs influence. (But hopefully better an better in term of opportunity cost, as production is less and less precious as time passes).
2a) Just make the upgraded units worse and worse as time advance. Technologies unlocking them give diplomatic buildings, so that they are still interesting.
2b) Discard the multiple diplomatic units. Just keep one units, which cost increase with time similarly to missionaries/faith buildings/...
As 2a was simpler to implement (mostly already done), and was working good enough, it was considered (last year) that there was no need for a deeper change.