Why are rows returned as stale when fetching TimeSeries data with multiple sequential queries on the same RowSet?

I’m using GridDB Cloud with the Python client (griddb_python) and found a reproducible issue where executing multiple sequential queries on a TimeSeries container causes the second query to return stale data unless I recreate the RowSet.

Schema

import griddb_python as griddb

conInfo = griddb.ContainerInfo(
    name="ts_metrics",
    column_info_list=[
        ("ts", griddb.Type.TIMESTAMP),
        ("cpu", griddb.Type.FLOAT)
    ],
    type=griddb.ContainerType.TIME_SERIES,
    row_key=True
)
store.put_container(conInfo, True)

Insert test data

import datetime

container = store.get_container("ts_metrics")

base = datetime.datetime(2024, 6, 1, 0, 0)
for i in range(5):
    container.put((base + datetime.timedelta(minutes=i), 0.5 + i))

Query 1 (works)

query = container.query("SELECT * WHERE ts >= TIMESTAMP('2024-06-01T00:00:00')")
rs = query.fetch(False)
print([row for row in rs])

Returns:

[(2024-06-01 00:00:00, 0.5), ..., (00:04:00, 4.5)]

Query 2 on the same RowSet (stale data)

query = container.query("SELECT * WHERE ts >= TIMESTAMP('2024-06-01T00:02:00')")
rs = query.fetch(False)
print([row for row in rs])

Expected: Rows from 00:02 onward.

Actual: It prints the results from Query 1, even though the SQL text changed.

If I recreate the container or create a new RowSet object manually, the problem disappears.

Question

Is there a known limitation in the GridDB Python client that prevents reusing a TimeSeries RowSet for sequential queries, causing stale-buffer reads, and is a new RowSet required for every query execution?

What I have tried

• Calling rs.close() and query.close() before reusing → no effect

• Calling container.flush() and container.commit() → no effect

• Recreating only the query → still stale

• Recreating the container object → fixes it

• Tested on GridDB CE 5.3 and GridDB Cloud → identical behavior

• TimeSeries-specific; Collections do not show this issue