Not sure if it applies to your situation, but I've found that very often mysterious failures disappear if you use an alternate 3.3v power source that you can be certain is capable of delivering ample current on demand, and then run wire from that directly to the points of large current consumption (of which the RFM69HW would certainly be one) rather than rely on just traces. I'd start with that, at least until after you get it working.
[Edit: Having said that, though, I don't believe I needed to do so on an "official" NodeMCU, which comes with an AMS1117-3.3V converter. If you're driving that from a good enough 5v USB source (say, a 500ma capable one, which is fairly typical), that should be sufficient, and IIRC it worked OK for me without needing an auxillary power source. I tried a number of the different off-the-shelf ESP8266 platforms, and I do recollect that not all of them came equipped with enough power to handle the current requirements of both the ESP8266 and the RFM69HW. For instance, at least in my case, the WeMo D1 Mini couldn't supply enough juice on its own.]