What I do is I add additional sounds as I’m writing/recording, but I don’t take any away until the final mix.
For example, if I get to a point where I think the guitar sound is no longer working, I will duplicate the guitar track and then modify the duplicate track’s sound until I get the new guitar sound that I want. I might mute the original guitar track at some point, or I might just let them both run together for a while until later deciding to mute one or the other.
Then when I’m done writing, I take a break before mixing. Ideally I wait until a different day. I start the mix by duplicating the writing session and turning all the faders down. Then I go track by track and bring up the drums, then bass, etc. while building the mix. If there are 2 (or even 3 or 4) guitar tracks, I can use my fresh mixing ears to either choose one of them to keep, or I can combine them. In some cases, I put one in the left channel and one in the right. But at the very least, I can refer back to the original guitar sound by muting the later tracks. Since everything is still there, there is no chance of accidentally deleting an awesome sound because my ears were tired when I was writing, or it got late and I wasn’t thinking as sharply. In some cases, I find that the original sound was actually the best and I throw away the newer tracks I made as I was writing. In some cases, I find that the original sound might not be as good as the later ones, but the original sound fit in with the rest of the mix better and added some kind of spark, so I end up being glad that I kept it.
Generally speaking, we have too many tracks today for most sessions. Even an iPad with a 64-bit chip can do 32 stereo tracks. So instead of thinking in terms of replacing, it is often better to think in terms of adding. Give yourself more options when you’re mixing and your view of the song is fresh and you are seeing the finished song.