Revisiting SSL for sound event detection: complementary fusion and adaptive post-processing