huh, on linux the sequential read was sucking much worse like this https://mastodon.cloud/@lamp/110835206975409950 https://mastodon.cloud/@lamp/110834052545026942 while I was cloning it.

maybe it has something to do with reading the first 1GiB vs the whole thing... which would make sense if it had SLC cache but I don't see any mention of SLC cache...