You didnot tell us the PVID of your VLAN configuration. So we cannot have full concept of your problem. Besides, How did you connect your network? Why you give ports for your WAN LAN? And do the WAN LAN and your LAN network completely isolated?

Seems the problem is related to ARP floods from the ISP provider.
After a while, the switch stops forwarding arp request.
There is another post in the forum about this

I've send a mail to the support, hopefully they can solve this

Boat loads of ARP's are usually the result of something wrong like a crashed switch, NOT the cause.

When dealing with heavy traffic the first thing we think of is to turn ON the "Storm control" and "Flow control" but both of these sick functions invariably corrupt the VLANs stacks by dropping packets.
So just make sure these 2 functions stay OFF.
Once your VLAN stacks are corrupted... you have to Delete your VLANs then redefine them 'cause rebooting cold is of no help.


EDIT: "QOS by port" is broken on SG108EV2 - While you are dealing with heavy traffic QOS *blocks/stops* ports on lower priorities.
I have tested the different combinations of mixing port priority assignments (greater, equal, lower)...
As I understand the "4 priorities stacks are emptied top down": currently if you're flooding the switch on a high priority port then the lesser priority ports are going to stall and drop frames.

The question is why TPL puts out gear without testing the chip reference design??
I am done pulling my hair with the SG108EV2 box. The only thing that works is to Reset the switch and keep it on default Vanilla config: VLAN1 only, no VLAN, no QOS... It's junk!