hi jcameron
it's always good to go back to basics if your voice network has gone crazy. If possible, set up a test SSID on one AP. Lock the virtual AP to "allowed-band a". Connect a test phone to that test SSID. Note the channel and the power the AP is on. Do some testing to observe the maximum coverage radius of this AP, the behavior of the client as it gets to the edge of coverage etc.
Once you have a feel for it, put the same test SSID on another AP that is "next" to the first AP, e.g. the first expected roaming target. Even better (from a testing perspective, not a wifi design perspective) if that AP is same channel and same tx power. Later you can force them to be on different channels same UNII and the introduce more channel spacing. If the second AP is not on the same channel, you will need to be aware of the channel setup of the i62 etc (as per the config guide). Do the same kind of testing now with these two APs. Do you get expected roaming behavior ? Where is the handover point when walking from AP1 to AP2 ? Make sure to try AP225 to AP225 and AP225 to AP335 (and vice versa).
Consider (for testing) removing various config optimisations from your testing virtual-ap and ssid profiles to help rule them out, e.g. if your roaming is still bad then get rid of local probe request threshold on the test SSID profile, and as above, you can even go as far as forcing the two test APs to be on the same channel with same tx power. Related note: make sure that you have reasonably consistent tx powers across all of your APs, don't let ARM set a huge range, voice clients do not generally like it and it makes for uneven roaming. Also make sure your AP channels are matching the expectation of the i62 config (e.g. 9 max, DFS or not).
Regarding the intermittent connectivity issue - if you find that the voice behaviour is good, then leave the test phone on the test SSID for a longer duration and see if it develops any issue. Again, start with one AP, then move to two etc. AP 2xx and AP3xx have different chipsets, there can be variation in behaviour regarding all the knobs in the SSID profile, so perhaps start to rule them out - indeed you could start with a clean SSID profile and layer in the optimisations one by one and observe the behaviour after each (the Ascom guide touches a few things that are not so commonly tweaked, you may have hit on a combo that has issues).
This is a lot of work, unfortunately voice issues often can be. I'll wrap up by saying that my immediate suspicions based on your symptoms are:
a) AP has no target to roam to, despite starting to look at -70 it doesn't find the next AP for some reason, I'd be looking at DFS/channel mismatches, power level imbalance etc.
b) the connectivity issues may turn out to be something at the AP, rather than an RF related issue. The above testing suggestions are designed to rule in our out RF as a trigger, once that is done, it's easy enough to setup a capture (be it wifi or packet level) to catch the device and AP interaction. As it stands right now there are too many variables to know where to look (or you get into this circle of collecting all the things)
hth
-jeff