I recently started upgrading our small business line to image 18.104.22.168. This includes an SG300-20 and five SG300-28P 's. The first thing I noticed is that the switch applies port power well before the switch has finished booting and is forwarding frames. Image 22.214.171.124 was perfect and waited to add power once the switch was operational on all ports. Definitely not the end of the world and possibly not a bug at all; our IP phones eventually time out and retry and the switch is then operational.
The second thing is that in layer 3 mode, setting a static IP on the managment VLAN address does correctly add 'no ip address dhcp' to the config file, but NOT the IP address and subnet mask. This must be entered manually from the CLI. I did a copy run start and rebooted a switch, and was left locked out with no choice but to use the console to add the missing information. This happened on two separate 28P 's.
On the second switch that had the missing ip address in VLAN 1 (but did indeed have the no ip address dhcp entry), only the first bank of ports had PoE. Removing and restoring AC power brought PoE back to all ports. Out of all strange behaviors, this one worried me the most.
The third problem I noticed was that LLDP no longer gave our Mitel 5330 IP phones VLAN information. Despite rebooting to defaults, resets, etc... Rolled back to image 126.96.36.199 and it suddenly worked again. I have since decided to use DHCP to provide that information to the phones, and turned off auto smartport as I don't like blinking Save all the time, and endless macros to scroll through in the config file. No matter what image you use, LLDP doesn't provide VLAN info on initial switch boot, after a power inline never / auto, and sometimes even after and port shutdown and power inline never/auto. The only way to get it for sure is to unplug and re-plug the phone. This was why I switched to DHCP for that task.
Dismissing these originally as upgrade config incompatibilities, I rebooted to factory defaults and reprogrammed manually. All these problems persisted. I'm currently running half equipment the new image 188.8.131.52 and half 184.108.40.206. It seems currently image 220.127.116.11. is the best image to use as it is stable and also does not have the MAC table port flooding bug that 18.104.22.168 and below had. 22.214.171.124 does have some nice new features, but I'm too scared to deploy it to switches over 10 miles away from the office.