[@Muttonhawk] I've done some maths.

Let us consider just the repayment time for rebuilding an existing Holy Site, not the added complexity of having Might left to make more Holy Sites. Suppose the original Holy Site was worth [i]x[/i] Might. You tear down the Holy Site and replace it with one worth [i]y[/i] Might. You thus have [i]y[/i] less Might, and your MPT is now [i]y[/i]/2 rather than [i]x[/i]/2. You break even once you reach Turn [i]t[/i], given by
[center][i]t x/[/i]2=[i]t y/[/i]2 - [i]y[/i][/center]
[center][i]t[/i] = 2 [i]y / [/i]([i]y[/i]-[i]x[/i])[/center]
For large [i]y[/i], this approaches 2, as for the repayment time of a new Holy Site.

For Capy's scenario with Arpeggio, an 8 Might Holy Site rebuilt for 32 Might, the repayment time will be 2 and 2/3 turns. For the Citadel, a 5 Might Holy Site, to break even within 3 Turns requires spending 15 Might. Spending 10 Might sees me break even in 4 Turns.

(Although, I need to consider things very carefully, since if I don't spend too much Might this Turn, I can finish Teknall's major project this Turn. Or I can wait until next Turn and not stress too much. But then I won't have the result of the project until later.)