How do I trust a drone's autonomy in hard conditions?

You use the autonomous features, but in the hard conditions you do not fully trust them, and that hesitation is worth listening to. The system was built and tested for the ordinary case. The hard conditions are exactly where its assumptions start to slip, and where your experience earns its keep. Trust here is not all-or-nothing; it is calibrated. In clean conditions, let the autonomy do its work and save your attention. When the wind, the light, the terrain, or the stakes turn against you, tighten your grip and stay ready to take over. The machine does not know when it is out of its depth. You do. That awareness of when to trust and when to intervene is the real skill now. So learn the edges of the system as carefully as you learned to fly. Note the exact conditions where it has failed or hesitated before, and treat those as your cue to take manual control early rather than late. Trust it where it is strong, and be the pilot where it is weak.