Corrupted data when Xarray writes to Zarr Datetime64 dtype

What happened?

Hi,
when using the new Zarr v3 DateTime64 dtype, I have trouble correctly writing to it with xarray - I have not found a way to write the correct values.
I believe it is probably related to some CF coding enabled when it should not be, or something like that.

Am I doing something wrong?

What did you expect to happen?

I expected Xarray to not corrupt data

Minimal Complete Verifiable Example

# /// script
# requires-python = ">=3.11"
# dependencies = [
#   "xarray[complete]@git+https://github.com/pydata/xarray.git@main",
# ]
# ///
#
# This script automatically imports the development branch of xarray to check for issues.
# Please delete this header if you have _not_ tested this script with `uv run`!

import xarray as xr
xr.show_versions()

import numpy as np
import zarr
import zarr.storage
import zarr.dtype

store = zarr.storage.MemoryStore()

g = zarr.create_group(store)
a = g.create_array('a', shape=(2,), dtype=zarr.dtype.DateTime64(unit='s', scale_factor=1), dimension_names=['time'])
a[:] = np.array(['2025-01-01', '2025-01-02'], dtype='<M8[us]')

ds = xr.open_zarr(store, chunks=None, consolidated=False).compute()
print(ds)  # reading is ok
# <xarray.Dataset> Size: 16B
# Dimensions:  (time: 2)
# Dimensions without coordinates: time
# Data variables:
#     a        (time) datetime64[s] 16B 2025-01-01 2025-01-02

ds.to_zarr(store, mode='r+', consolidated=False)

ds = xr.open_zarr(store, chunks=None, consolidated=False).compute()
print(ds)  # writing writes bogus values
# <xarray.Dataset> Size: 16B
# Dimensions:  (time: 2)
# Dimensions without coordinates: time
# Data variables:
#     a        (time) datetime64[s] 16B 1970-01-01 1970-01-01T00:00:01

Steps to reproduce

uv run the script

MVCE confirmation

• Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
• Complete example — the example is self-contained, including all data and the text of any traceback.
• Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
• New issue — a search of GitHub Issues suggests this is not a duplicate.
• Recent environment — the issue occurs with the latest version of xarray and its dependencies.

Relevant log output

<xarray.Dataset> Size: 16B
Dimensions:  (time: 2)
Dimensions without coordinates: time
Data variables:
    a        (time) datetime64[s] 16B 2025-01-01 2025-01-02
<xarray.Dataset> Size: 16B
Dimensions:  (time: 2)
Dimensions without coordinates: time
Data variables:
    a        (time) datetime64[s] 16B 1970-01-01 1970-01-01T00:00:01

Anything else we need to know?

No response

Environment

Start by running the provided MVCE with the listed xarray, NumPy, and Zarr versions, then trace the open_zarr and to_zarr paths for datetime64 handling and CF encoding. The fix is complete when the round trip preserves the original 2025-01-01 and 2025-01-02 values instead of writing epoch-based values.