Host IronPython instance per TPL thread?

[vadimkantorov]

Is it possible to host IronPython interpreter per worker thread from the TPL library? Should it be possible to somehow override the TPL's ThreadPool to ask them to initialize and keep a IronPython runtime object in threadlocal state?

As I just found out, the cPython unfortunately does not support sub-interpreters and isolated per-thread interpreters just yet (my shallow investigation in pythonnet/pythonnet#2170), so I wonder if IronPython can support in my migration of TPL pipeline using C# payloads to TPL pipeline using Python payloads (I understand the limitations of IronPython / unavailability of NumPy/PyTorch and so forth)