My main question: if you want to keep your relays running this way, how would you generally know the relay was accepting transactions at all?
I considered this as a somewhat drastic measure for one of our two relays. Unlike the other, it has peers who like to submit about 12 copies of transactions. The first in each of those groups succeeds, and of course all the following copies fail with invalid UTxO’s:
So it certainly keeps the load average down on the affected relay to turn off TraceMempool… presumably by saving the CPU load of having to mathematically invalidate the remaining ~11 transactions… but then unfortunately we have no confirmation the original, valid transaction is going through either. 