Also updated the docstring for Annotations.orig_time and added a RuntimeWarning when a string is provided but it gets converted to None. Thoughts on this?

The code for annotations works fine, but changing _convert_times() has some knock on effects for converting data objects to dataframes.

This is an alternative approach to fix the current errors:

elif time_format == "datetime":
    # make ISO8601 (microsecond) compatible
    times = (to_timedelta(times + first_time, unit="s") + meas_date).to_numpy()
    times = DatetimeIndex(
        [time.replace(nanosecond=0) for time in times], tz=times[0].tz
    )

but I'm not 100% on how others want to proceed so I'll pause for now.

Hey @drammock, please could I get your input on the proposed changes. The original annotations issue is resolved, but there are some implementation details that weren't discussed in the issue that I'd like to get suggestions on.
That would also help with deciding how I can resolve the issues with dataframe conversions.

The code for annotations works fine, but changing _convert_times() has some knock on effects for converting data objects to dataframes.

This is an alternative approach to fix the current errors:

elif time_format == "datetime":
    # make ISO8601 (microsecond) compatible
    times = (to_timedelta(times + first_time, unit="s") + meas_date).to_numpy()
    times = DatetimeIndex(
        [time.replace(nanosecond=0) for time in times], tz=times[0].tz
    )

but I'm not 100% on how others want to proceed so I'll pause for now.

Turns out this doesn't fully resolve things. Once the proposed changes to _convert_times() are better established I can tackle this properly.