1602 and 1702 APs are mostly connected via Cisco SB SF300 non PoE switches wit PWR-INJ4/ PWR-INJ5/ CISCO SB SB-PWR-INJ2 as power source.
After upgrading controller software from 8.3.133 to 8.3.140 my only 2802i AP stopped work and boot began to fail with the message like this:
From the non PoE switches side I see different pictures depending on the model. On Cisco SB SF300 ports are down, on catalyst 3750V2 port is UP but no mac-address present. If I connect same AP2802i via PoE catalyst 4500 Ap2802I starts to work.
So I spouse there is a bug in 8.3.140. It affects only 2802 and 3802 connected via PoE injector.
So what should I do now? Fall back to 8.3.133 or wait to 8.3.’?’ update? We have no budget to upgrade all branch switches to PoE model.
We have no service contract, but depending on this bug purchasing of 100 2802 AP can be delayed or canceled.
The version you go with is up to you. 18.104.22.168 had some issues that made me want to get away from it. I wasn't surprised, it was a rushed release due to the WPA2 security vulnerabilities and fixes. If you didn't have any issues with it, then maybe downgrading is the way to go for you. Honestly, I'd try the TAC route and get a fixed version of 8.3.140.x, but that's just my opinion.
Regarding the other switch, I don't think it will make a difference what the access switch is, whether PoE or not. With my situation 100M ports didn't work under any circumstances, I had to use a 1G port to get the 2802 APs online with that software. If you end up using a PoE injector on a PoE switch make sure to disable PoE on the port, otherwise CDP advertises different PoE info and the AP can get confused. This is unrelated to the software issue. I had a 2802 AP on a 30W power injector that wouldn't turn the radios on because the PoE switch was advertising only 15.4 watts.