In Civ2 and Civ3, we had to (and still have to) do micromanagment to reduce overflow, since beakers/research or city production that was more than was needed was lost. They addressed that in Civ4 and later games, by allowing overflow to carry over into the next tech or item in the city build queue. Indeed, the Civ4 players used that as a point of optimization; they would get several items 1 turn from completion, shuffle the order, to end with a big overflow that could complete multiple units in a single turn. Civ5, BERT, and Civ6 changed the way the city queue works. Instead of paying gold to *finish* what's being built, the player pays gold (or faith) to build something else *entirely*.
Holding back on era score to avoid overflow / wastage feels like the same problem -- maybe with the same solution? Or would the snowball effect be too overpowered?