NCCL timeouts when reading or writing HF checkpoints for 70B Llama model during multinode training

**Describe the bug:**
There are two related bugs:
1. When I start multinode traning from a HF checkpoint of a Llama 70B model, it fails because of an nccl 600 seconds timeout when loading the checkpoint
2. When I finish multinode training of Llama 70B, a similar error hapens when saving a HF checkpoint.

An example of config and logs is attached.

[ws_64.6ca44755_logs.zip](https://github.com/user-attachments/files/24672886/ws_64.6ca44755_logs.zip)

**Describe how to reproduce:**
I encountered it during OMT training. A minimal reproduction example has to be figured out.

**Describe the expected behavior:**
No errors

**Environment:**
At the very least, specify the versions of fairseq2, PyTorch, Python, and CUDA along with your operating system and, if relevant, GPU model.
RSC or AWS-SC, python3.12, fairseq2==0.6, torch 2.8.0, CUDA 12.8

**Additional Context:**
Add any other context about the bug here.

Start with the attached ws_64.6ca44755_logs.zip and the example configuration from the OMT training run. Investigate the NCCL timeout during Hugging Face checkpoint loading and saving in multinode training, then establish a minimal reproduction. Done means both operations complete without NCCL timeout errors.

Written by the indexing model from the issue text.