I imagine this has been going on for a while but it's only really starting to bother me, but I have an issue with infinitely long textgen using SillyTavern and Ollama. The way I usually work things is I have my responses set to streaming so I'll watch it gen and then hit stop somewhere where I feel like it's getting off track and I'll trim it, then write a response to steer it or just have the AI create responses off of what it just said and reroll those until I get something good.
If I just hit send and let the AI fire off while I walk away it will never stop. I just did so when I started making this message and pic related just kept going when I wanted only a 60 token response, getting stuck in a death spiral repeating the same sentence over and over again. It doesn't always death spiral like that though, usually it doesn't get stuck in a hard loop like that but it will just keep generating story. Just fired another one off and since I just found the token counter extension I can check, it gave me a response that 1625 tokens and would still be continuing if I didn't stop it.
Addendum, I don't have a deep understanding of tuning the Sillytavern settings, I just switch between presets and mess with things to try and get a more varied response when I reroll outside of the AI just giving me more or less the same message again, occasionally toggling Instruct mode on and off to see if anything seems like it changes.