Let me describe the issue. The client has 2 ZD3000 series located somewhere in Antwerp Belgium. The customer has several sites and access points are installed at these remote locations (around 270 access points in total). The access points are a mix of ZF7372, R500, T300 and ZF2741). We have 2 sites where there are only ZF2741 access points installed. On Sunday, the client performed an upgrade from 188.8.131.52 build 15 to 184.108.40.206 Build 9, nothing else was changed to the config. This is when problems started to arise, especially at the sites with ZF2741 access points installed only. The affected client devices are all clients that connect to a ZF2741 using 220.127.116.11 build 9 (so not only scanners but laptops also). I see that the clients can authenticate - associate - 4-way handshake - receive an ip address and can send data (normal data, QOS data) and go to sleep. Nevertheless we see a lot of packet loss (around 85%) We see this if we do a ping from the ZD to the clients and from the APs to the clients. My best guess is that there is a new feature in 9.7 that causes these issues (like bandbalancing/steering/client load balancing?). If we rollback to 9.6 everything works just fine.
Nothing in ZD code should impact your performance like that. Tech support would ask to see the AP suupport info files from your 2741s when in use on both versions. Do you use streamlined authentication, ie WPA2 (not WPA-Mixed) with AES (not Auto TKIP+AES)? Assuming your scanners do WPA2-PSK? I've seen issues when using the mixed/auto combination in certain clients. Otw, a ticket with Tech support might be helpful.
Actually, there was internal software library changes between 9.6.x and 9.7 that increased AP load on processor and memory resources.
You can run 2942/2741 along with 7962/7363/7341s on 18.104.22.168.12. But consider reducing number of WLANs <= 5, no more than one tunneled back to the ZD, have a fast path to RADIUS if using 802.1x, and you should be ok.