M4300 12x12f no vlans while stacking failover

If you're using embedded DHCP server pool for each VLAN in your M4300 core, and if your edge S3300 switches are directly connected to it with corrent VLAN proliferation between edge and core, then yes you should be good to go without DHCP Relay agents for your VLANs.

The number of 10G stacking links between two M4300 switches at the core, or two S3300 switches at the edge, will all depend how much traffic is expected East-West between the switches in each stack. At a minimum I would always advise 2 x 10G links for stacking.

LAG with or without LACP is always a requirement when attaching one Ethernet device with several links to a stack of several switches (one link per switch in your case) for load-balancing and redundancy. Without LAG, you are creating a loop since the stack is meant to be one device in the network.

I don't seem to fully understand your testing procedure when connecting your PCs and triggering failover / failback at your core, but losing one ping or two doesn't matter, especially when ICMP not prioritized. Ethernet protocol handles this and there is no service interruption with broken TCP/UDP sessions. If you are losing 5 pings, you might want to understand why. Again, if LACP correctly configured in your LAGs between your edge switches stack and your core stack, there must be sub-second failover at your core with M4300.

Last, in order to understand the difference between a simple Stacking Member, and a Stacking Member handling the Opr Standby = Backup Management Role in M4300 stack, please consult the M4300 Software Administration manual starting page 434 (chapter 21: Switch Stacks). I hope this will greatly helps!

Regards,