Should every slot have a block?

There has been some internal discussion that Ouroboros needs and recommends [CITATION NEEDED] empty slots. (not producing a block every slot) Specifically that this is needed to heal short term forks.

This is counter to other internal discussions suggesting that we want to produce lots of blocks. (at /least/ one per slot, maybe more.) This is supposed to aid in stake being offline and help ensure robust transaction processing.

I’d like to try to resolve these two approaches and better understand how we should tune things.

More blocks?
Less blocks?

Quoting from the Praos paper:

In order to cope with the ∆-semisynchronous setting we introduce the concept of “empty slots” which occur with sufficient frequency to enable shor tperiods of silence that facilitate synchronization. This feature of the protocol gives also its moniker, “Praos”, meaning “mellow”, or “gentle”.

Later on in section 4 it comes into play as influencing D^f_{Z,A}, which is related to D^f_α
On page 19 there is this:

Remark. Intuitively, the theorem asserts that sampling the characteristic string in the ∆-semi-synchronous setting with protocol parameter f according to D^f_α is, for the purpose of analyzing divergence, comparable to the synchronous setting in which the honest stake has been reduced from α to α(1−f)^∆.

1-f is the proportion of empty blocks. Our ∆ is 2, f is 0.5 in theory but our current code for that is buggy.