I'm attempting to convert the flutter/plugin integration tests from Intel to ARM host machines (PR), but they are often timing out. These same tests are currently running on Intel hosts, where the do not have this problem, but we want them to be on ARM going forward.

I originally thought that they were randomly hanging, but looking closer it seems like the problem is that they are actually just running out of time, even though the 60 minute timeout should be plenty (the Intel versions of these tests generally take about 20-35 minutes per shard, and that's with 4 shards; we increased to 5 as part of the ARM transition since that's getting to be long).

It looks like all of the Xcode steps are just taking a lot longer. For instance, compare this Intel run of shard 4/4 for a PR that built all packages (the plugins repo only runs tests on changed packages most of the time, but things like rolls or CI changes run all packages) to this ARM run for shard 5/5 on an all-packages PR. They are running the same set of packages (video_player, video_player_avfoundation, webview_flutter, webview_flutter_wkwebview), except that the Intel one is also running url_launcher. Comparing some step times:

Build:         Intel 6.2m, ARM 25m
Xcode analyze: Intel 1.6m, ARM 9m
Native test:   Intel 6.1m, ARM 12m

I would expect the ARM machines to be as fast or faster, not 2-4x slower. Are these machines spec'd substantially differently?