For Question One: The time bucket is primarily used for further calculations based on minute data synthesized by the TS engine. It's important to note the timing for closing the window: for a left-open right-closed window, the window closes when the first received timestamp is greater than or equal to the right boundary; for a left-closed right-open window, it closes when the first received timestamp is greater than or equal to the right boundary timestamp minus one. For example, for the window [09:00, 09:05), the window closes when data with a timestamp greater than or equal to 09:04 is received.

For the window from 21m to 24m, which is left-closed and right-open, the actual window corresponds to 21m-23m. After aligning the data precision, the corresponding timestamps for the window are 02:21:00.000 to 02:23:00.000. Therefore, data with timestamps exceeding 23:00.000 and falling between 24:00:00.000 will not be included in the calculations. It can be seen that the time bucket requires the data precision and the window to be as consistent as possible; otherwise, data may be lost.

For Question Two: The reason for the unexpected results is that the first data timestamp has been adjusted based on step and round time. Please refer to the manual for the time series database for clarification. Since version 2.00.14.4, the daily time series engine supports outputting the first window based on session begin, rather than the timestamp adjusted from the first data.