It's
getnetworkhashps but it's only in the 0.16.3 client (don't use it for txs) and was transcribed from
the corresponding code in the unreleased Primecoin branch "primecoin-0.16.3". Its accuracy in terms of representing hashpower is questionable but its representation of
relative variation should be a bit more reliable. I don't know why CrossVerse didn't include it in the 0.15.99 code, maybe it's profoundly unreliable.
You are correct in your observation that it is based on an inference: “The hashrate can be calculated from the expected rate of finding a block (1440 a day), the actual rate of finding a block and the current difficulty”.
I reworked the plot in my own client to show a full week's worth (10080 blocks), the plot shows that difficulty is still adjusting to the new network hash rate:

Cheers
Graham