Random Nonce vs Incremental Nonce - in a pool

Ok this is an aged old question since the beginning of mining

https://www.reddit.com/r/Bitcoin/comments/3uyo0q/random_nonce_vs_incremental_nonce/

I tried playing with some codes, and as it turns out, it is possible to do 'random nonce' while mining in a pool.

Normally, the miner iterates the nonces adding 1 for each new hash. And as to how 'random', would depend on how one tweak the algorithms, I tried 'random increments', i.e. randomly skipping increments, and it seemed pretty feasible. mining pools accepts the hashes per normal if they match the difficulty requirements.

Has anyone else tried this, and what are your observations?

ok, i'd contribute my 2 cents:

- it may after all be useful for 'slow' miners / computers. modern gpus and asics may be able to iterate the whole 4 billion numbers in a second, or some may be lesser these days, I'd guess. But the average 'slow' computers are everywhere and won't be able to iterate even a fraction of that 4 billion hashes before the block is found and the next block starts.

- the rate of valid shares discovered apparently is quite similar between linear increments vs random increments. It isn't necessary slower one way or the other.

- the volatility in hit rates seemed higher between linear increments vs random increments. it seemed to keep testing for shares for stretches without finding shares, then in a sweet spot, finds a whole bundle of shares in rapid succession and submit them. I think this would probably happen in linear increments as well, but it seemed less observed.

- random increments means skipping 'slots' i.e. leaving lots of voids unexplored, it is probably ok for a 'slow' computer as after all it won't complete with the 4 billion iteration before the next block turns up.

- there are a lot of unknowns as in that would random increments find the higher difficulty shares and potentially solve the block? this is something very difficult to tell. i.e. linear vs random, which is better? ;)