difference between Wireshark and application-level UDP read

You have probably realised that you are starting to go around in circles. Wireshark uses pcap https://en.wikipedia.org/wiki/Pcap[^] which is a very low level packet sniffing API. It accesses the hardware directly and bypasses the TCP stack altogether. This makes it useful for diagnosing network traffic and network hardware. If your are on Windows it is WinPcap http://www.winpcap.org/[^]

In your case it is able to confirm packets are arriving. Is it more robust than XYZ. No. It is a specific tool and once you know the network is working who cares.

When you added threads to your program probably the best thing you could have added was moving from UDP to TCP. This gives you guarantees, retransmits, handshaking etc. If you then added your own packet IDs and maybe checksums you would not only be getting closer to a reliable system but also have some means of verification.

Whether to use TCP or UDP would have to be based on proper analysis but generally the use of UDP means that you are prepared for lost packets. If lost data is an issue then TCP is the way.

It would however still be down to you to get it all working and that can also be a challenge.

Quote:

I observed that Wireshark capture is more robust (?) than C-application with CSocket (ReceiveFrom()) to receive UDP data.

Not more robust. If implemented correctly, your application should get every single packet wireshark gets.

Quote:

What factors make the difference? Is Wireshark doing some low-level tasks?

Yes it is, as already mentioned, it uses a low level driver to set your network card in "promiscuous mode" and capture anything that passes through it.

Quote:

Can I use the capture of Wireshark as a comparison to assess my UDP receiving application?

Sure, that's what the application is used for after all... packet analysis. If your application is not receiving something, you can analyze the packets in wireshark to see if there is something wrong with them. For example, if some higher-level fields are not being filled in correctly... you may never see it at the application level. That's on purpose, you wouldn't want to capture a bunch of junk on the wire in your application would you? If there is nothing wrong with the packets and you're still losing them, that would point towards architectural problems with your software design. Perhaps you may be spending too much time doing other tasks before servicing your sockets and your buffers are filling up and dropping data. Either way, wireshark is your friend for analysis of network traffic.