Are you required to declare past conflicts of interest?

This question is triggered by the recent allegations about Willie Soon failing to declare conflicts of interest in his papers on climate change. It got me wondering about situations where an obvious conflict of interest existed in the past, but doesn't exists anymore at the time where a paper is published.

Are there generally accepted practices on when to declare past conflicts of interest, or is this entirely up to each journal? Requiring disclosure of decades-old conflicts of interests does feel a bit silly to me, but making them disappear if they just happened shortly before one worked on a paper seems rather problematic as well.