I believe that this conflict between the Math.SE community and the SO overlords (we have had puzzlement, confrontation, petitioning, and exasperation) regarding the commenting system stems from an inherent difference between the two camps' conception of what the purpose of the commenting system is.
According to the SO design philosophy (or usage guide-- whatever it is that the FAQ represents), the purpose of comments is to identify ambiguous portions of questions (with the hope that the OP will make the question more precise), and also to identify errors, i.e. reasons to downvote, an existing answer.
It is my impression that Math.SE users that find the enter-key behavior in comments odious do so because they expect (or wish) for the comments they leave to have substantive content, i.e. they wish that the comments function not simply to identify errors or ambiguities in questions and answers, but to enrich the content of said questions and answers by providing further detail and perspective. This, if I read the FAQ correctly, is morally a misuse of the commenting system: it seems that by design of the SO software (of the SO overlords), any substantive addition to the site ought to be in the form of either a newly-posed answer or an edit to an existing one.
Now I may be misinterpreting the FAQ (English being my second language and all and this being the internet), but I believe that the larger question then becomes a private question, namely, whether the discussion of mathematics that the SO platform provides (as outlined in the FAQ) is a platform under which you yourself wish to discuss mathematics.
My personal belief is that the moral misuse of the commenting system by Math.SE to leave substantive comments (as opposed to the technical misuse of leaving multi-line comments) is something inherent to the discussion of mathematics owing to the high density of information inherent in mathematical statements (that is, what qualifies as a 'simple' clarification does not simultaneously qualify as a 'short' clarification because of the levels of abstraction that have to be identified and navigated through).
In any case, I believe that making the comment-box behave like a chat-box (which is exactly the reasoning behind the new behavior) is detrimental to the quality of the comments (on any .SE, not just Math.) exactly because until it reduces them to chat and hence to insubstantial contributions to the content of the discussion. More importantly, the moral requirement that substantive contributions be in the form of full-blown answers imposes a linear structure on all contributions, and life, or at least mathematics, is too interconnected for any one answer to effectively (and artificially) arrange all relevant information in a beginning-middle-end form (a phenomenon not unrelated to why it's so difficult to write a good textbook or even learn mathematics in the first place); there will always be strands of an answer of significant interest that are worthy of being addressed but simply do not cohere, expositionally, with all the other information in the answer, thus rendering editing infeasible.
For these reasons, I believe that the implicit policy of what comments ought to be like, which I take to be the one documented in the FAQ and which is the one that underlies the soundness (in the eyes of SO overlords) of having the enter-key submit comments, is fundamentally flawed and antithetical to the goal of .SE sites to function as effective Q&A forums. Furthermore, it is the implicit policy about what comments ought to be that I perceive as the reason behind the recent discord here at Meta.Math.SE.