22163 miners with 51.5MH/s from nicehash[0] and 26395 miners with 935.9MH/s from miningpoolstats[1] are different and therefore can be summarized, right?

[0] web.archive.org/web/20191224132957i…m/main/api/v2/public/orders/active2

[1] web.archive.org/web/20191224141157i…poolstats.stream/data/monero.js?t=1

And a disquieting number of times where the instantaneous block rate is pretty high.

Small coins decided to abondon PoW and switched PoS some time ago. They didn't find any solution for constantly changing hashrate and delay between blocks. Their super-duper difficulty algos didn't has anything except beatuful name.

Monero isn't a small coin on the market of randomx mining but it may be changed.

i wonder if the tiny lag introduced by epoch changes can cause cascading effects in diff adjustment ....

You can write a simulation and test it.

no cascading effect, but it will decrease the difficulty by about 0.001%

because nobody is hashing for about 4 seconds every 2048 blocks

The source of "effects in diff adjustment" is obviously not constant hashrate of miners. But there is no solution unless you want to speed up mining reward distribution by lowering diff after dropping hashrate.

the effect is so small that it can be ignored

No, I mean this effect "the instantaneous block rate is pretty high"

"instantaneous rate" is a bit hard to measure in a Poisson process