Running media_unittests under ARM64 I see:

...
[00008.731] netstack: convSockOpt: TODO IPPROTO_IP optname=1
[00153.098] pkgsvr: 2018/07/19 12:46:10 pkgfs: importing package from "c6600142c9297029bae1e1dd82dde8388121f64df38d4b56ac0884b38f07d549"
[00153.470] pkgsvr: 2018/07/19 12:46:10 pkgfs: blob already present for {media_unittests 0}: "aecb120a2a0dc8adcc581078e78035282499a02c8babcfb05d8473222039e2e1"
[00153.474] pkgsvr: 2018/07/19 12:46:10 pkgfs: blob already present for {media_unittests 0}: "2dd45dd32d0cca2518e66729952360d090568924c6b762d0a0b7a0e3b5243c63"
[00153.479] pkgsvr: 2018/07/19 12:46:10 pkgfs: blob already present for {media_unittests 0}: "0240ee13cde720f449720da18916991466303f728caedef50c573f012471ed3c"
[00153.486] pkgsvr: 2018/07/19 12:46:10 pkgfs: skipping bad package entry: ""
[00153.495] pkgsvr: 2018/07/19 12:46:10 pkgfs: asking amber to fetch 333 blobs for {media_unittests 0}
[00173.313] pkgsvr: 2018/07/19 12:46:30 pkgfs: index: Amber notification error: ErrPeerClosed: zx.Channel.Write
[00173.313] pkgsvr: 2018/07/19 12:46:30 pkgfs: index: Amber notification error: ErrPeerClosed: zx.Channel.Write
...

This suggests that we are taking more than two minutes to push the binary into the VM, before we even get around to inserting it into the blob store.