Will a valid PDF 1.x document be a valid PDF 2.0 document? Will PDF 1.x readers gracefully accept PDF 2.0 documents?
Those who worked on PDF 2.0 in ISO made great effort to be as compatible with present-day implementations as possible and to break as little as we could possibly do. Some changes (like adding UTF-8 string object support) weren't possible in a means that is totally backward-compatible. Most changes, however, are clarifications to the standard and updates to better align the ISO standard to the behaviors of implementations. For example, the section on Tagged PDF was rewritten in a way that is much clearer and easier to understand, and includes a new tag set for PDF 2.0 that does not match the PDF 1.7 tag set. However, the PDF 2.0 tag set is not the default - the default remains the PDF 1.7 tag namespace - so that Tagged PDF files can be promoted to PDF 2.0 without the tags needing to be rewritten.
PDF 2.0 is an evolution of the PDF standard, not a drastic change to the specification. It does include some new capabilities - such as the ability to specify string objects in UTF-8 format - that PDF 1.7 viewers will not be able to handle correctly. However, most PDF 1.7 viewers should handle most PDF 2.0 files perfectly well. In this respect, your comparison to PDF 1.6 vs. PDF 1.7 is a fair comparison: viewers that don't know about PDF 2.0 revisions or features will either ignore these or not handle these correctly. However, I expect most PDF 2.0 files will be written in a way where they are compatible with present-day PDF 1.7 viewing implementations.
If you're interested in more information, Peter Wyatt's presentation at PDF Days Europe 2017 can be downloaded from https://www.pdfa.org/wp-content/uploads/2017/05/PDF20Keynote.pdf ; or you can watch it via https://youtu.be/AR4lGlDh9Ac.
Disclaimer: I am the chairman of the PDF Association