Everything works fine until I disconnect PC2 from the network or it gets powered off. After that, every device in my network will receive the packet from our Embedded board while their IP addresses are different from PC2 for about 1 minute and after that, the Embedded board tries to find the network for 192.168.124.10 IP with ARP packets.
That is to be expected. While you use off-topic network devices, the basic effect is universal.
When the destination node's switch port goes down, its MAC address is removed from the switch's MAC address table. That in turn causes any frames with that destination address to be flooded to all ports (but the ingress one), mimicking a bus network or repeater. Eventually, the cached ARP entry on the source device times out and failing a subsequent ARP resolution, the transmission should stop.
That is normal behavior of Ethernet 'by design'.
It is our main problem because all devices receive our Embedded system data and the network gets fully interrupted by the high level of received data.
That seems to be a problem with the devices in your network - normally, a device (ie. its NIC) ignores frames with a destination address that is neither its own (hardcoded or programmed) MAC address nor a broadcast/multicast address.
While flooding wastes bandwidth in your network, there shouldn't be any real impact - unless the wasted bandwidth significantly approaches or even exceeds the overall network capability. In that cause you should seriously consider limiting the relevant broadcast domain, ie. contain the problem source in a new VLAN and IP subnet that you can then route to the intended destination.