How many deals is enough before the diversification actually does something?

Correlation caps it. Once your positions all depend on cap rates at exit and on rent growth holding up, adding the 30th deal removes sponsor-specific risk and does nothing about the shared risk, which is the bigger term. I'd rather have 12 positions spread across property type, geography, vintage year and debt versus equity than 40 multifamily value-adds all bought in the same window. Vintage spreading is the one people skip, and it's the cheapest to do since you just deploy over time instead of all at once.

Also, the $10k minimum is far from universal. Plenty of the debt offerings go down to $1k or less, which is how position count gets built without $400k. The returns are lower, and it is a different risk to hold alongside the equity.

There is no single number, and one point deserves more weight than it usually gets. A debt position and an equity position in the same building are not really two positions. The property either performs or it does not. Diversifying by capital stack inside one deal counts for less than it looks like it does, so a position count that includes those pairs overstates the spread.