Presumably this depends on your perceptions of:
- The time and resources required to make the major changes
- The probability of acceptance should you make the major changes
- The degree to which the paper will be improved by the changes and the value you place on that improvement
- The probability of acceptance in another journal were you to not make the changes (or just the minor changes), and how much work any future journal might request from you as part of making changes, and any cost assigned to the delay in publication that this may involve
- How much lower impact the alternate journals are (and how much you care about impact factors and rankings)?
If I get a revise and resubmit from a high ranking journal, I would almost always try to make the changes requested. There is a little bit of luck involved with publishing in good journals. If you go on to submit elsewhere, these other journals may reject, or they may have equally onerous edits. However, once you get a revise and resubmit, if you do a good job of implementing the requested changes, then the editor will typically accept the manuscript. Furthermore, at least in my world, publishing in prestigious high-impact journals has many broader benefits (e.g., increased exposure for your idea, and improved cv which helps for grants, promotions, job security, etc.).
That said, I can think of a few exceptions where I might not make the changes.
The editor says something about liking the broad topic, but is basically requesting a whole new study (e.g., collecting a whole new dataset with a bigger sample, better measures, different design, etc.). In particular, this is often combined by comments suggesting that the editor would be happy to consider such a study: i.e., not really offering a commitment to publish even if such a study was implemented well.
Sometimes, I'm working on a paper that is not my core focus: e.g., perhaps converting a student thesis for publication, or helping out a colleague. The paper might be fine as is. I understand that it could be even better were we to collect whole new datasets or invest substantial time in some other respect. But in such a case, this time and these resources may feel like a distraction from my core program of research.
Ultimately, it's up to you to weigh the issues up. But I would assign a high-value to publishing in high impact journals and taking advantage of the opportunity that you have been given to make revisions.