While this implementation is likely correct and probably the thing we should do, I worry about the performance hit this will take. Do we have a benchmark test to see how much of an impact this change has? Perhaps we need to also refactor things a bit to make sure we only call the appendRaw method for big chunks of data where it's worth syncing and not for every single entry for example.

Do we have a benchmark test to see how much of an impact this change has?

Yes, you're right! It may cause overhead of syncing unless a given benchmark proves otherwise 👍

Perhaps we need to also refactor things a bit to make sure we only call the appendRaw method for big chunks of data where it's worth syncing and not for every single entry for example.

I will look into this. Thanks for sharing!