'open_mfdataset' zarr zip timestamp issue

What happened?

We have been collecting Satellite data and we save each image as one {time}.zarr.zip file.
We then collate the images using xr.open_mfdataset and same them to large.zarr.zip file.
When loading this file the timestamps are all the same.

This bug did not appear in 2022.3.0 but it did in 2022.6.0

I tried to keep this as minimum as possible, but its a bit of a long example. Hopefully the comments help.

Sorry if this has already been reported, but I could not find it in the issue list

What did you expect to happen?

Expected the time stamps to reflect the data that went in

Minimal Complete Verifiable Example

import pandas as pd
import xarray as xr
import numpy as np
from datetime import datetime, timedelta
import zarr
import os
import glob

# ids and times
path = "tmp.zarr.zip"
ids = np.array(range(0, 10))
times = [datetime(2022, 9, 1) + timedelta(minutes=60 * i) for i in range(0, 10)]


# make 10 random zipp files
for time in times:
    dataset = xr.DataArray(
        np.random.uniform(size=(1, len(ids))),
        coords=(("time", [time]), ("id", ids)),
        name="data",
    ).to_dataset(name="data")

    file_name = f"tmp_dir/{time.isoformat()}.zarr.zip"

    if os.path.exists(file_name):
        os.remove(file_name)
    with zarr.ZipStore(file_name) as store:
        dataset.to_zarr(store)

# load them all together
files = list(glob.glob(f"tmp_dir/*.zarr.zip"))
dataset = xr.open_mfdataset(files, engine="zarr").sortby("time")

# this is fine!
assert pd.to_datetime(dataset.time.values[0]) == times[0]
assert pd.to_datetime(dataset.time.values[1]) == times[1]

# save to file
if os.path.exists(path):
    os.remove(path)
with zarr.ZipStore(path) as store:
    dataset.to_zarr(store)

# read the file
dataset_read = xr.open_dataset(path, engine="zarr")
print(dataset_read)

# this casues an error
assert pd.to_datetime(dataset_read.time.values[0]) == times[0]
assert pd.to_datetime(dataset_read.time.values[1]) == times[1]

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.

Relevant log output

/Users/peterdudfield/Documents/Github/nwp/venv/lib/python3.8/site-packages/xarray/core/dataset.py:2060: SerializationWarning: saving variable None with floating point data as an integer dtype without any _FillValue to use for NaNs
  return to_zarr(  # type: ignore
<xarray.Dataset>
Dimensions:  (time: 10, id: 10)
Coordinates:
  * id       (id) int64 0 1 2 3 4 5 6 7 8 9
  * time     (time) datetime64[ns] 2022-09-01 2022-09-01 ... 2022-09-01
Data variables:
    data     (time, id) float64 ...
Traceback (most recent call last):
  File "/Users/peterdudfield/Documents/Github/nwp/venv/lib/python3.8/site-packages/IPython/core/interactiveshell.py", line 3251, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-16-45f86e8a5977>", line 36, in <module>
    assert pd.to_datetime(dataset_read.time.values[1]) == times[1]
AssertionError

Anything else we need to know?

No response

Environment

Reproduce the provided example using xarray.open_mfdataset, Dataset.to_zarr, and xarray.open_dataset, comparing the reported 2022.3.0 and 2022.6.0 behavior. Trace the Zarr timestamp serialization path used when writing the combined dataset, then verify that the final dataset preserves all input timestamps and passes the example assertions.