Hi Felix
Thanks for you reply!
When i refered before the possibility of being VSWR, it is because if the antenna is near some 'ground' it could represent a much lower impedance to the transmitter and cause an high VSWR. I have no ground plane but have inside a small enclosure the NodeMCU, the RFM69HW, an 16x02LCD + I2C adapter and a protoboard connecting everything, so the antenna will be always near any of these modules.
I'm more inclined to your suggestion of being an EMI problem in the SPI / power connection between the ESP and the RFM, that might be somehow prone to the interference of the antenna signal.
The communication is working well between 2 moteinos, and also between a Moteino and the ESP+RFM, if I use a coil antenna o the ESP+RFM side (this is what was used until now and it was working, but the signal was weak also due to the bad placement of the coil antenna inside the box, so I decided to test alternative approaches and got into this issue). I can observe with a third moteino working as a sniffer that when this problem occurs no message is really transmitted from the ESP+RFM node. I just don't find a reason to it, because even if it is an EMI problem, I think that the RFM only turns on the transmitter after having received all the message in its FIFO, so it should transmit it independently of the SPI communication with the ESP, that will only occur after turning on the transmitter, don't you think? And also it worked well in a breadboard, with much longer wires. Now it was assembled in a protoboard, the wires are now shorter, but there is no ground plane because I haven't made a PCB for this node yet. One thing that I thought trying was to create some ground plane using cooper shielding tape, but didn't tried it yet. I have used tens of RFM69 modules without any issue until now, so I'm still a bit surprised with this issue. It is true that never tried to put the antenna inside the box, I have always used external SMA antennas or simple 1/4 wave straight wires.
Best regards
Fernando