One observation is - I would not fix something that is not broken.
In real cases, congestion becomes an issue with a lot more nodes than just a few.
Also - if you have [1 listener] and [1 transmitter at 1 time] on 1 "frequency" (if you use LoRa then its already spread spectrum), then I don't see the real gain of this scheme. It's not like the gateway can listen on 1 frequency, and transmit on another, at the same time. It cannot be in 2 places at 1 time.
Whoever transmits, normally needs to:
1) Switch to the transmitting channel
2) Sense energy in the channel first to detect an ongoing transmission - rather than just shoot in the dark
3) Transmit
What that transmission channel is makes no difference, the congestion happens in that channel, and the gateway is always tied up in it. A node transmitting in the TX channel, while the gateway is gone to RX channel ... not sure if it can be called congestion reduction, but it sure doesn't help the node get its message to the gateway.
Switching frequencies and transmission modes involves delays so.
I see this as merely an extra effort of switching frequencies, rather than keeping the same frequency and managing transmissions properly and implement a retry mechanism.
I would keep the channels at least a few hundred khz apart for LoRa modulation, 500khz or even 1mhz won't hurt.