Interpretation of time-averaged diagnostic fields

Fields in contain diagnostic fields, the output of which is controlled by a diag_table. For variables where time_avg = T in GFS_diagnostics.F90, how should one interpret the vertical levels of the output?

Is it correct to assume that the fields are accumulated on each level, even though the pressure at each level is changing in time? And that the time averaging is taking that accumulated value and simply dividing by the number of seconds in the time-averaging interval? If so, the "reference pressure" in the same file is having the same thing done to it?

Or, is something more complex being done behind the scenes such that diagnostics are interpolated to set pressure levels during accumulation?

Thanks for any help,