Is it a bug? Read the description ⬇️

I uploaded a LoRA yesterday for training, but so far it is not ready. It alternates between 8% and 33%. Apart from the times when he returns to the initial state of waiting in line, it now seems to have stabilized and the progress of the LoRA does not leave the place. This is very strange, it shouldn't take more than a day for a LoRA to be ready, I don't know what I do, and the worst thing is that it was a LoRA for the DiT.2 model, the most expensive it has, I feel like I spent credits now... Another thing I would like to share is that I feel that the models are kind of weak or failing more often. I tried to use the Tsubaki 1,1 model because a combo of LoRAs that I like to use is only available in this model, but the results are not coming out as before, the quality is terrible and the results present many problems even though I clearly describe what I want.