Ark Kernel Fails to Connect

I am encountering an issue where the Ark kernel fails to establish a connection within a Windows-based Docker container running JupyterLab. I am able to successfully run R code using IRkernel in the same environment, but Ark consistently exhibits connection failures.

Environment

Docker Version: 28.1.1
Base Image: mcr.microsoft.com/windows/server:ltsc2022

JupyterLab Version: 4.4.3
Conda Version: py311_25.5.1-0

R Version: 4.5.1
Ark Version: 0.1.195
IRkernel Version: 1.3.2

Logs

[W 2025-06-30 15:49:02.525 ServerApp] Nudge: attempt 120 on kernel 339b791d-abc8-4cfa-8db3-4e1aab3d613d
[E 2025-06-30 15:49:02.844 ServerApp] Uncaught exception GET /api/kernels/339b791d-abc8-4cfa-8db3-4e1aab3d613d/channels?session_id=233900e2-03a1-47d3-aae5-f4bc8ab0630b (10.14.52.63)
    HTTPServerRequest(protocol='http', host='<host>:1234', method='GET', uri='/api/kernels/339b791d-abc8-4cfa-8db3-4e1aab3d613d/channels?session_id=233900e2-03a1-47d3-aae5-f4bc8ab0630b', version='HTTP/1.1', remote_ip='10.14.52.63')
    Traceback (most recent call last):
      File "C:\Miniconda3\Lib\site-packages\tornado\websocket.py", line 967, in _accept_connection
        await open_result
      File "C:\Miniconda3\Lib\site-packages\jupyter_server\services\kernels\websocket.py", line 75, in open
        await self.connection.connect()
    TimeoutError: Timeout
[W 2025-06-30 15:49:02.868 ServerApp] Timeout waiting for kernel_info reply from 339b791d-abc8-4cfa-8db3-4e1aab3d613d
[I 2025-06-30 15:49:02.869 ServerApp] Connecting to kernel 339b791d-abc8-4cfa-8db3-4e1aab3d613d.
  2025-06-30T22:49:02.874131Z ERROR  Received subscription message, but no `subscription_tx` is available to confirm on. Have we already received a subscription message once before?
    at crates\amalthea\src\socket\iopub.rs:283

[W 2025-06-30 15:49:02.913 ServerApp] Replacing stale connection: 339b791d-abc8-4cfa-8db3-4e1aab3d613d:233900e2-03a1-47d3-aae5-f4bc8ab0630b
[W 2025-06-30 15:49:07.414 ServerApp] Nudge: attempt 10 on kernel 339b791d-abc8-4cfa-8db3-4e1aab3d613d

... repeats

Steps to Reproduce

FROM mcr.microsoft.com/windows/server:ltsc2022

# Install miniconda
RUN powershell Invoke-WebRequest -Uri https://repo.anaconda.com/miniconda/Miniconda3-py311_25.5.1-0-Windows-x86_64.exe -OutFile MinicondaInstaller.exe
RUN powershell Start-Process -FilePath MinicondaInstaller.exe -ArgumentList '/InstallationType=AllUsers','/RegisterPython=0','/S','/D=C:\Miniconda3' -NoNewWindow -Wait
RUN powershell Remove-Item -Force MinicondaInstaller.exe
RUN powershell -Command "[Environment]::SetEnvironmentVariable('Path', $env:Path + ';C:\Miniconda3;C:\Miniconda3\Scripts;C:\Miniconda3\condabin', 'Machine')"

RUN conda install jupyterlab -y
RUN ipython profile create

# Install R
RUN powershell Invoke-WebRequest -OutFile R-4.5.1-win.exe https://cran.r-project.org/bin/windows/base/R-4.5.1-win.exe
RUN R-4.5.1-win.exe /SILENT /DIR="C:\Program Files\R\R-4.5.1" && powershell Remove-Item "R-4.5.1-win.exe"
RUN powershell -Command "[Environment]::SetEnvironmentVariable('Path', $env:Path + ';C:\Program Files\R\R-4.5.1\bin', 'User')"
RUN powershell -Command "[Environment]::SetEnvironmentVariable('R_HOME', 'C:\Program Files\R\R-4.5.1', 'User')"

# Install Ark
RUN powershell Invoke-WebRequest -OutFile ark.zip https://github.com/posit-dev/ark/releases/download/0.1.195/ark-0.1.195-windows-x64.zip
RUN powershell Expand-Archive -Path "ark.zip" -DestinationPath "C:\ark" && powershell Remove-Item "ark.zip"
RUN powershell -Command "[Environment]::SetEnvironmentVariable('Path', $env:Path + ';C:\ark', 'User')"
RUN ark --install

# Install IRkernel
RUN Rscript --no-save --no-restore -e "install.packages('IRkernel', repos='https://cloud.r-project.org'); IRkernel::installspec()"

ENTRYPOINT ["jupyter", "lab", "--ip=0.0.0.0", "--port=1234", "--allow-root"]

1. Build Image

docker build -t <name> Dockerfile .

2. Run container

docker run --rm -p 1234:1234 -it <name>

3. Open URL and select Ark kernel.

4. Attempt to run any R code.

cc @josiahparry

Reproduce the failure using the provided Dockerfile and Ark 0.1.195 setup, then inspect the subscription handling at crates\amalthea\src\socket\iopub.rs. Compare the Ark kernel connection behavior with the working IRkernel setup and trace the repeated timeout. Done means Ark connects in the Windows container and can execute R code without the kernel_info or subscription errors.