Let me start with the statement that I work for Aruba, so my experience is based mainly on the Aruba ARM solution, and this may bias my personal opinion as well.
I found that most of my customers actively run the ARM (Adaptive Radio Management) that comes in the Aruba solution in the default settings. If you have a good coverage, network, that works pretty well. Lets take out the exceptions for very high density like crowd coverage in stadiums, event halls, and very hostile RF environments where you typically either tune ARM or revert back to static channel assignments.
Also, I learned that many other vendors in the wireless space, without calling specific names, advise against automatic radio management for several reasons that seem mostly related to their implementation. It either does convert quickly to a stable channel plan, or it does just a bad job, does not scale, does not adapt to changes in the environment because it only scans and calibrates during nights so adapts to a situation without users on the network. Please also note that if you are planning statically you will need to know your RF environment very well; do spectrum analysis on a regular schedule, check channel utilization, which require well trained engineers.
The reason to use ARM in most Aruba deployments is that typically channel assignment by the ARM algorithm has a much quicker outcome and is near or probably better than manual channel picking. So it saves time and if your channel plan is not better by ARM it probably is not much worse. Also they way Aruba implemented ARM is that if suddenly a source of interference, both WiFi and non WiFi, comes up, or the RF environment changes because your users move and change the RF environment, that a static channel plan will not respond to that; so you stick to the static plan that you developed. Also on 5 GHz if you use DFS channels, and you probably do, if a radar event is detected with a static channel assignment the radio has no oter choice than to go offline, potentially creating coverage holes.
In some networks, it may work to help ARM and remove some channels on APs were you see radar detections. Also if you have clients that refuse to work on DFS channels, it may help to create a basic grid with the non-DFS channels and let the other APs select a DFS channel for additional capacity for clients (most clients!) that do support DFS.
And yes, there are exceptions.... as mentioned before. It is known to me that different WLAN specialists have different views on it, either theoretical or pragmatic. I do believe in theory you can create a better channel plan through solid engineering; on the other hand I'm a pragmatist and believe that good is good enough and if the easy and automated way does not provide the desired results you can always revert back to the manual engineering work.
Hope this personal view helps. May be others can let know what their experience is?
Herman