This game is optimized really poorly, and loads whenever it likes. An SSD seems to perform better, as you might expect.

If we were to count times with and without loads (specifically when the mace spins in the bottom-right) everyone's time would improve, and the gaps between times would decrease as well.

Of course, this is Garshasp we're talking about. It's not that important in the grand scheme of things.

1) I'll have you know that garshasp is a really good game

We have four runs on the board with minutes between them; once discrepancies in loading times become the discerning factor between runs, maybe it will be worth it. Even then, though, loading times (other than game start) aren't very high in variance or even that frequent, right? It seems like more clerical work to track this without much gain.

Which loading zones are you concerned about? Here are the ones that come to mind:
Game start — Sometimes a couple seconds, sometimes more. If you get a bad load here, you can reset. A lot of the weird stuff here happens outside of mace-spin
Goblin roofs — If you do the walk-off strat here, you hit a several-second loading zone. There's a perfectly good way to avoid it, which is to jump on the rest of the roofs in a deev-approved fashion
Yella fella skip — A few seconds when you jump while the game loads under you, but the time here feels pretty consistent (anecdotally)

It would definitely level the playing field. However, like Stacky said, the runs are minutes apart, not seconds, and the loading times, though varying, largely seem to stay somewhat consistent. Right now, it would be more work than it's worth to track loading times. I'm not totally against it, but it doesn't seem necessary yet.