My guess as to why operator algebras gets a bad rap: In my mind there are certain areas which mathematicians are inherently drawn to, chief among them number theory, topology/geometry, and any area which has a physical interpretation/application (pure analysis, PDE, etc). Mathematicians also tend to like problems which have a low barrier of entry, and so appreciate them even if they don't immediately contribute to understanding the big areas.
Operator algebras sits at the other polar extreme. There is a high barrier of entry (functional analysis, algebra, algebraic topology, geometry, physics) which would be forgivable if it had led to the solution of deep problems in one of the big areas. Unfortunately, the big problems it has contributed to haven't been finished off in an entirely satisfactory way (I'll explain some examples below). Part of this is attributable to the attitudes of those working in the field. For example, much of the work in noncommutative geometry is about finding the right notion of a "quantum geometry" which can involve quite a bit of wheedling with definitions and exploring examples. Many papers are about giving good definitions, the kind of bedrock stuff that's been established in e.g. topology/geometry for decades now. And even simple examples can be difficult; it's still not understood how to appropriately extend Riemannian geometry to the noncommutative torus.
So the field is still young, requires a high entry barrier, and the impetus to break that barrier isn't high enough for those outside the field because there aren't big name fundamental problems that operator algebras have solved. I know this is somewhat reductive, but it's my best attempt for now.
Finally some examples of the successes of operator algebras (please correct me if you see mistakes here or have additions):
All recent results on the Novikov Conjecture that I know of use operator algebras in an essential way, and often prove the stronger Baum-Connes Conjecture (see e.g. work of Gennadi Kasparov and Vincent Lafforgue). At this point, we know that the Baum-Connes Conjecture holds for many classes of groups (e.g. hyperbolic, amenable, and even some property-T groups).
Elliott's program for classifying C^*-algebras using K-theoretic invariants has found wide applications in classifying dynamical systems. I get the sense that dynamicists often rephrase these K-theoretic invariants in more familiar dynamical language, which can obscure the influence of operator algebras, but I believe there are some cases where you really do need to use the language of K-theory to fully describe classification results. Here are two recent papers: http://arxiv.org/abs/1406.2382, http://arxiv.org/abs/1502.06658. Generally one can look for work of George Elliott, N. Christopher Phillips and papers by Giordano, Putnam, and Skau.
I should also add that there is a large body of work on quasicrystals and the gap labeling conjecture of Jean Bellissard where operator algebras have played a major role. References are easily found on Bellissard's website.
The work of Packer/Rieffel/Luef shows many connections between frame theory and operator algebras. On a related note, some of the widest reaching results on the HRT Conjecture were proven by Linnell using operator algebraic methods (see http://arxiv.org/abs/math/9807057). This result is in a similar spirit to the zero-divisor conjecture for group rings, which can also be approached using operator algebraic methods (see Luck's work on L^2-torsion and this paper http://arxiv.org/abs/1202.1213).
There's also Connes' work on foliations, but unfortunately I don't know enough to describe that in any detail.