In `@plugwise_usb/nodes/circle.py`:
- Around line 885-891: The month overflow handling around calculating
circle_timestamp is wrong: when (difference := days_diff - days_to_end_of_month)
> 0 you adjust corrected_day but do not advance the month, producing incorrect
dates; replace the manual day correction logic (the block using
last_day_of_month, days_to_end_of_month, corrected_day and
dt_now.replace(day=...)) by computing circle_timestamp as dt_now +
timedelta(days=days_diff) (using the existing dt_now and days_diff variables) so
month/year boundaries are handled correctly and remove the fragile
corrected_day/replace logic.

In `@tests/test_usb.py`:
- Around line 3043-3083: The test_test_clock_synchronize_month_overflow case
doesn't force a month rollover because the frozen date "2026-01-31" is already
Saturday (weekday 5) and mock_clock_get_send sets response.day_of_week.value = 5
so days_diff == 0; update the test so the mocked day_of_week differs by +1 to
force crossing into February (e.g., set response.day_of_week.value = 6) or
change the frozen time in the `@freeze_time` decorator to a Friday date so that
mock_clock_get_send's day_of_week=5 results in a +1 day rollover; modify the
mock_clock_get_send implementation (and/or the freeze_time value) inside
test_clock_synchronize_month_overflow to ensure days_diff != 0 and the code path
that adds timedelta(days=1) is exercised.

In `@plugwise_usb/nodes/circle.py`:
- Around line 912-913: The docstring for async def _send_clock_set_req currently
incorrectly references "CirclePlusRealTimeClockSetRequest"; update it to
accurately describe that this method sends a CircleClockSetRequest (e.g., "Send
CircleClockSetRequest.") so the docstring matches the implemented behavior in
_send_clock_set_req and avoids confusion with CirclePlus methods.
- Around line 883-888: The early return calls _send_clock_set_req() when
dt_now.weekday() != response.day_of_week.value but the null-check for
self._node_protocols happens later, which can cause an AttributeError; either
move the existing self._node_protocols is None check so it runs before the early
return or add the same guard at the start of _send_clock_set_req() (and
return/raise gracefully) so _send_clock_set_req() never accesses
self._node_protocols.max when _node_protocols is None; update the call site or
the helper accordingly and ensure logging/behavior remains consistent.