XML 1.0 5th edition
(1) It's true that 5th edition doesn't do anything to change namespace names from URIs to IRIs ("internationalised" URIs). But that can be done separately, in a revision to Namespaces 1.0. In any case, XML namespace names are usually just treated as strings, and most software doesn't check that they are valid URIs. I don't really see the problem: even if namespace names remain URIs, that's a minor issue compared with being able to use more characters in element and attribute names. There's no "consistency" between element names and namespace names that needs to be maintaned. And typically a namespace name just appears once in a declaration at the start of a document: it doesn't affect readability in the way that element and attribute names do.
(2) IBM didn't push through "features" to suit mainframe users - they just persuaded the W3C to support the traditional IBM NEL line-end character in XML 1.1. Why anyone else should care about this is beyond me.
(3) The main problem with XML 1.1 is that Microsoft doesn't support it. One reason for that may be that they didn't manage to get support for arbitrary binary data in XML documents; XML is supposed to be a textual format.
(4) All the major XML software producers - including Microsoft - are likely to support the 5th edition changes, so the compatibility problem will be small. It's not as if zillions of XML vocabularies using the new characters are going to appear overnight.