plotly 6.3.0 does not work with numpy datetime

[StageTired]

I upgraded my requirements from plotly==5.9.0 to plotly==6.3.0 and observed a regression. Markers on the plot disappeared without any error.

Simply put, my code takes a figure as input and plots a dot at the end of each line. Simple reproducible example:

import numpy as np
import pandas as pd
import plotly.graph_objs as pgo
import plotly.subplots

np.random.seed(42)

# random data
dates = pd.date_range(start="2025-01-01", periods=3, freq="D")
data = np.random.randn(3, 1)
df = pd.DataFrame(data, index=dates, columns=["value"])

# this is plotted elsewhere
fig = plotly.subplots.make_subplots()

for column in df.columns:
    trace = pgo.Scatter(
        x=df.index,
        y=df[column],
        mode="lines",
    )

    fig.add_trace(trace)

# take the figure and
# render last point of each line using fig.data
for i, trace in enumerate(fig.data):
    input_x = trace.x[-1]
    input_y = trace.y[-1]
    print("Input x:", input_x, type(input_x))
    print("Input y:", input_y, type(input_y))

    fig.add_scatter(
        x=(input_x,),
        y=(input_y,),
        mode="markers+text",
    )
    print(f"Plotted x value:", fig.data[-1].x)

fig.show()

If I run this with different plotly versions I get different type for the timestamp:

With plotly<6, the figure looks like this

With plotly>=6, it looks like this

I am aware that I can convert numpy datetime back to datetime.datetime to fix my problem but why does plotly treat numpy datetime as a unix timestamp? It's not being converted to an actual date and there is no error either.
Simply put I would expect x axis to be a date and not an impossibly large number:

import numpy as np
import plotly.subplots

fig = plotly.subplots.make_subplots()

x = np.datetime64('2025-09-26T01:00:00.000000000')
y = 1.23

fig.add_scatter(
    x=(x,),
    y=(y,),
    mode="markers+text",
)
print(f"Plotted x value:", fig.data[-1].x)

fig.show()