First keep in mind that the old MX2 papers were designed to be unfinishable. Marks in that range back then were essentially sufficient for scraping away band E4. Some things like Q8 of 2001 are still what I'd consider insane difficulty for a typical 4U student.
So here's where we begin talking about exam strategy. Out of your three points, only 2. Focus is something that I'd continue to reinforce.
"Speed" can be a huge trap for a maths student. Of course, at this point into the HSC speed is definitely something you'd want. But it's not the biggest priority. The problem with speed is that some people think they need to speed up with what they struggle at. Trying to achieve this very quickly is almost impossible. For questions you're good at, sometimes speed makes you more prone to silly mistakes. For questions in topics you struggle at, the word "speed" can be a big source for pressure and crumbling down.
When it comes to speeding up, you should only speed up at things you've become fully confident in. Those are things where you have reasonably low likelihoods of making silly mistakes in, for example a classic calculate \( \frac{\overline{z}}{w} \) type question, or \( \int xe^x\,dx \). In particular, with these questions that you have good speed in, do them first. Forget that the paper is split into 8 questions (or rather 6 in your actual exam) - just blast the easy shit out of the way and enjoy the fact you've basically secured those marks.
A note on silly mistakes: You should be compiling a list of all the silly mistakes you make, and look at that list
1. twice before going to sleep
2. twice before starting a new math paper
so that in the exam, you will inflict minor panic about if you made that silly mistake. After you receive the minor panic, scan through your work and see if it's happened (NOT properly check yet, just SCAN). If it doesn't look obvious, whatever, come back to it later. If it does, well, if you're correct then congratulations. Otherwise, you can thank yourself for knowing what silly mistake it was you had to look out for.
And of course, remember, once you start giving up that is a good sign that you should start checking your entire paper. Rope yourself into checking it yourself first, before looking at the answers/solutions.
Then, the issue of "hard questions". At this stage, I believe now that you've seen the rough end of the spectrum, you should now go for the softer parts. This means you should now start at the 2017 paper, and build your way back to somewhere (say, 2013 or 2009). This lets you tackle papers that have the same format as your current paper (i.e. 10 M/C and only 6 long responses), and more likely the same difficulty (because trust me, some of the earlier papers were hard). You might think that going all out with hard questions is a good idea, and to be fair it can be. But it's a double edged sword in that if you don't play it out well, it can be demoralising. That is one thing you don't want to have, because that will start killing your motivation to study.
You have also identified that you struggle with conics and mechanics. To be fair, most textbook questions don't offer the same type of difficulty/style in conics/mechanics questions as in past papers, so going back to textbooks for that is hard. But given that you now know where your difficulties are, you now know what questions to skip in the paper. You should be doing every other question that you can do first, before looking at these.
Also, mechanics/conics questions are, like you say, often appearing in multiple parts. You should typically not try to do the questions in order. Reason being more often than not, you should be able to scab marks in a later part, by just assuming something given to you in the previous part. Another opportunity at mark maximisation.
And finally, make proper use of worked solutions. Look at why a certain method worked, not just how to do it.