I'm not seeing a lot of comments describing first-hand experience with a holacratic system, so I'll jump in.
I was a developer at Moveline (which shuttered, sadly, this past December). Around the start of 2014, the company adopted holacracy as its organizational model; while there were certainly other valid reasons to do so, but I know that it was in part motivated by the fact that another Las Vegas-based company, Zappos, was also an adherent. Over the course of the year, in addition to writing code, I also served as the elected secretary for the dev team and facilitator of the GCC (General Company Circle; top-most level of the hierarchy), attended holacracy training courses, and coached people throughout the company about the mechanics of holacracy and how to make it work for them. Suffice it to say, I was pretty immersed, so I feel like I can represent how things played out at Moveline fairly accurately.
The up-front costs associated with shifting from a traditional structure to any other structure can be pretty massive; holacracy is no exception to that rule. Bigwigs from the official holacracy company (HolacracyOne) came onsite to give talks and run the first several meetings, which cost a decent chunk of change as well as taking time out of a busy startup's day. Teams and roles needed to be defined using the terminology and mechanics of holacracy. There were impacts here and there to scheduling, to communication systems, and to getting things done in general. Eventually, however, the ramp-up phase ended and we were able to experience holacracy in full.
As you might expect, there were established figures in the company who felt strong resistance to holacracy. Some of them held significant managerial power before the transition; others wanted a flatter, less formal approach. Those tensions came up fairly frequently in meetings and other conversations, and it created an atmosphere of not being quite satisfied with how things were being run (not that these people should have stifled their opinions, mind you). This problem was compounded when we entered a major growth phase, going from under 40 employees to over 100 in the span of a few months. New hires were unfamiliar with holacracy, necessitating additional training, but just as importantly they entered an environment where low-grade dissatisfaction with holacracy was palpable. If there's a lesson here, it's that adopting holacracy requires far more buy-in than you might expect. (For what it's worth, I'm not aware that anyone ever felt unable to voice their dissatisfaction at Moveline; we had a culture of open and honest talk, but identifying an issue isn't the same as solving it.)
A natural question to ask at this point is "What about backing out of holacracy?", to which I'd reply that it's easier said than done. Since a not-insignificant amount of effort goes into the initial setup, it's easy to fall prey to the sunk cost fallacy. The rationalization is reinforced by HolacracyOne, who clearly stated to us that most companies won't really be fully operational as holacratic organizations for at least six months. Actually, the tenets of holacracy itself reinforce the idea that you just have to keep pushing forward no matter how weird it may feel: the first setup step effectively tears down a bunch of preconceptions about corporate structure, and you won't really feel like you've gotten to the point where you can critique holacracy until you rebuild that structure.
The flexibility you get for achieving the new structure is touted as a feature, but the less marketable truth is that it's a double-edged sword. Holacracy is purposefully vague on where the authority for hiring, firing, and budgets will live. Lead Links aren't managers, but they are accountable for resource allocation that isn't defined elsewhere — and it's up to your company to figure out what limits are placed, who places the limits and how, what constitutes a "resource", and so forth. It's entirely possible for an employee to have zero roles — and it's up to your company to figure out whether that has any correlation to terminating employment, changing teams, hiring needs, and so forth. Furthermore, self-defining teams present an unfortunate opportunity for organizational thrashing; at Moveline, we saw major revisions to the structure around meta-org work (HR, facilities, finance, etc.) and our sales/support teams every few months. Those upheavals, while well-intentioned, degraded individual performance, disrupted internal communication, and fed back into the underlying dissatisfaction.
I haven't read The Tyranny of Structurelessness cited in many other comments, but from the title alone I am assuming that it's an accurate depiction.
That said, there is one thing that I feel like holacracy helped with significantly: the emphasis on well-defined decision-making. When teams stopped saying "We will get this done" and started routing tasks to specific individuals (who had an obligation to perform those tasks according to the roles they held), it was clear that the reduced ambiguity had a positive impact. The use of "we" has its place in an organization, of course, but "we" is just a code word for "not sure who" when it comes to planning specific, concrete work that needs to get done. If you're not careful, things can get a bit political when it comes to defining authority (at Moveline, this often related to holacratic "domains"), but I'd still consider the technique of specific task attribution as something I'd want to bring to future workplaces.
I'm happy to answer any other questions about my views on holacracy that HNers might have.