Paused -> stealth mode on....

[genbasura]

I have been thinking like.. shouldn´t a "paused" account, try to avoid requests to OpenAI as much as possible? I mean, you never know what MLs they have on their end, if they may detect too many concurrent requests by ip or something... That may make accounts to be flagged... Also the refreshing of the tokens, I think it should happen if possible when the other accounts are not being used, you could have a time period since last time any account was used, and after X time, then refresh token or something like that... the idea is to have like 10 accounts, but 5-6 paused...

Also, at some point it may be good idea to be able to join multiple codex-lb instances, so they can run on different IP....

[svk27]

This may solve this issue, what do you think?
#330