OK, so for starters - there is no actual USB demodulation flow-graph in gr-satnogs. This means that the default flow-graph (which is FM) will get picked instead. What you’re hearing on that observation is pretty usual for FM-demodulated noise. Turn the squelch off on any FM receiver and you’ll get the same thing.
Even when using a preamplifer, setting the gain to nothing is a bad idea. You are, in effect, deafening your receiver. You need to have some level of gain set in the RTLSDR regardless of whether you have a preamp or not. If you have too little gain set, your SDR will be ‘internally noise limited’ - that is, you are being limited by the sensitivity of the SDR. What your really want is for your receive system to be just barely externally noise limited. This means you want just enough gain that the noise from the local environment (and there will be noise…) is just above the noise floor of your receive system.
You can test this by (while looking at your RF spectrum with something like GQRX) terminating the input of the LNA with a 50 ohm load, then unplug the load and plug in the antenna. If you see the noise floor rise just slightly, then you have your gain set about right. If you see no significant noise floor increase, then wind the gain up. If the noise floor rises a lot, then maybe wind back the gain a bit. You don’t want to have too much gain set, otherwise you may end up generating intermodulation products.