@Alexandre Girard good question — and honestly a really smart way to approach it for a harvested energy setup.
From what I've seen, the degradation is generally graceful rather than cliff-like, at least down to around -6dBm or -8dBm at moderate distances. The STM32WB07's receiver sensitivity is doing a lot of the heavy lifting on the link budget, so you have more headroom to play with on the TX side than you might expect. Where things tend to get ugly is when you're already pushing distance *and* you start stacking environmental factors (multipath, humidity, obstructions) on top of a reduced TX level — that's where the margin disappears faster than the math suggests.
For a harvested energy node at short-to-mid range, I'd honestly start your testing around -4dBm to -6dBm and measure actual RSSI variance over a few days rather than just a point-in-time test. The firmware repo has the TX power configuration exposed — worth pulling that and doing a sweep at your actual deployment distance to find where your PER starts climbing before you commit to a power budget number.
What's the rough link distance you're working with on these nodes? That'll change the answer a bit.
