Can a USB port passively 'listen only'?
USB CDC serial is quite a complicated protocol. Even if you are only sending serial commands one way, there is USB traffic going backwards and forwards all the time. Even when the virtual serial bus is quiet, the usb host is continuously asking the usb device "do you have anything to send" and the device says "no".
Consider sending the phrase "hi" to an arduino, and getting "hello" back. On the USB bus, it will look something like this:
Host: Do you have a message?
Device: No.
Host: I have a message for you
Device: OK
Host: hi\n
Device: OK
Host: Do you have a message?
Device: No.
Host: Do you have a message?
Device: hel
Host: Do you have a message?
Device: llo\n
Host: Do you have a message?
Device: No.
etc. etc.
These messages will also be mingled in with messages sent to other devices on the same hub (even if you don't have a USB hub in the system, there may well be one inside the computer. So you could easily see messages to your mouse and keyboard mixed in).
The USB protocol is extremely complicated, so splitting it in the way you mention is not going to be practical. You can "sniff" it though. If you connect the D+ and D- lines, it is possible to see the traffic on the bus. As long as the sniffer doesn't try to manipulate the D+/D- lines, the USB bus will keep working. The sniffer would then need to identify which messages were relevant and decode them. Building a sniffer out of an arduino would be difficult but probably not impossible. It almost certainly isn't the best solution for your problem.
No you can't split USB, but what prevents you from connecting the UART receive pins of two Arduinos together so both receive same UART data, or making the first Arduino to send commands it does not understand to the other Arduino?
Can I use USB passively, as a listener only, with the second controller?
Technically yes, you can build a device that is passively looking at USB traffic and be able to get information from it and take any actions. In fact, devices like that do exist. They are called "USB Protocol Analyzers".
Typical architecture of such devices consists of an non-invasive sniffer on D+/D- lines (high-impedance voltage divider plus gain compensating wideband amplifier), which feed a standard (stand-alone) USB PHY. The serial-parallel output interface from the PHY (ULPI or UTMI or PIPE3) is then watched/logged by a sizeable FPGA that has decoding tools and can trigger off USB protocol events. Examples: Teledyne-LeCroy, Ellisys.
So you can watch USB traffic and get data. To get them in real time you will need to program FPGA and run it at sufficient clock rate to implement most elements of USB Serial Interface Engine (except forming ACK/NAK replies).