Non-reproducible results with num_workers=0

Nobody has claimed this yet.

### Bug description

Hello,
I have a question related to obtaining reproducible results when setting `num_workers` in a `torch.DataLoader` and using `pl.LightningDataModule` and `pl.Trainer`.

So far I am experiencing the following: when i set `num_workers=0` the results are different to those obtained when `num_workers>0` for any epoch other than the first (and my understanding is that the way the data is reshuffled within each epoch is somehow the same for all `num_workers>0` but different for `num_workers=0`). Is this an expected behaviour? Any suggestion to prevent it?

If you think this is a bug, I can provide a concise example that reproduces what I see (which however requires stripping out code from an internal library so I'd rather do it only if you think what I get is likely a bug and not an expected behaviour 😄 )

Thanks!

### What version are you seeing the problem on?

v2.5

### How to reproduce the bug

See code in the comment below

### Error messages and logs

No real error, just results are not reproducible for different `num_workers`

### Environment

Current environment

```
#- PyTorch Lightning Version (e.g., 2.5.0): 2.5.0
#- PyTorch Version (e.g., 2.5): 2.6.0
#- Python version (e.g., 3.12): 3.12.2
#- OS (e.g., Linux): Linux
#- CUDA/cuDNN version: running on CPU
#- GPU models and configuration: -
#- How you installed Lightning(`conda`, `pip`, source): conda
```

### More info

_No response_

cc @lantiga

Start with the reproduction code referenced in the comment and review how torch.DataLoader, pl.LightningDataModule, and pl.Trainer handle shuffling and worker counts. Compare results across epochs for num_workers=0 and num_workers>0, then establish whether the difference is expected or identify the behavior that needs correction.

Written by the indexing model from the issue text.