The revision in the field

I’ll run `context_verify` now to confirm the new bench tests reproduce FP-251 before applying the fix.
TESTS FAILED: built test_frontpanel (0 s), ran 1 test(s) (0 s)
0% tests passed, 1 tests failed out of 1
tests/test_frontpanel.c:39: CHECK_EQ(volume_get(), 23) failed: 17 != 23
tests/test_frontpanel.c:54: CHECK_EQ(freq_get(), 1000 + (i + 1) * 100) failed: 900 != 1100
not built: 6 other test program(s); none of them uses the code you changed
The new test reproduced the reported result exactly: three clockwise clicks changed volume from 20 to 17. The fix is in place; I’ll use `context_verify` for rev C and `cpp_test_select` to run the bench, encoder, and key tests in the rev B build.
PASS: built test_frontpanel, test_encoder_qei, test_knob (0 s), ran 3 test(s) (0 s)
100% tests passed, 0 tests failed out of 3
not built: 4 other test program(s); none of them uses the code you changed
diff --git a/drivers/encoder/encoder_qei.c b/drivers/encoder/encoder_qei.c
-    /* Reverse in hardware: inverting TI1 flips the count direction at no cost. */
-    hal_tim_encoder_init(cfg->tim, 0xFFFF, cfg->reverse);
+    /* Keep raw timer counts: encoder_detents() applies cfg->reverse for both drivers. */
+    hal_tim_encoder_init(cfg->tim, 0xFFFF, false);