Dear authors,

I have one major comment, similar to the major comment of referee #1, but potentially in another direction. This is the following point:

Lines 290 to 319: As I see it, the choice of the RT solver or spectral integration scheme has only a minor impact on simulated cloud properties. A better way to compare the differences would be to run the same simulation with a slightly different realization of the initial temperature field (superposed with some noise). This would lead to variations relative to the original simulation as well and would provide a basis to judge whether a difference caused by different RT methods in within or without the noise.
How is the effect on radiation quantities like div F?

Additionally, there are some technical slang and similar problems in the manuscript, which is not good. Please improve the respective text pieces thoroughly:

Line 38/39: Explain "atmospheric aggregation".

Lines 44-47: how can model errors affect cloud formation? Please rewrite.

Lines 50-53: Please rewrite this as well. As it stands, it is the coupling that is expensive. But it is the execution of an LES coupled to 3D radiative transfer that is expensive.

Line 62: The logic of your argumentation would be clearer with "even more expensive LBL calculations".

Line 78: The error is uncorrelated to what? And what does "The error is statistical" mean?

Line 85: please rewrite "... not meaningful to calculate a spectral band in one column and a different band in a neighbouring column."

Lines 166-169: please rewrite. As it stands, the coupling is solved instead of the equation system.

Line 171: The expansion of PETSc should be moved from ll 189/90 to here where it appears first.

Lines 183-185: An equation system can not be written as one matrix. A matrix is not an equation. Please rewrite.

Line 197: More efficient than what? Line 198: easier than what?

Line 227: What is a sparsity pattern? Please reformulate.

Lines 239/240: There is a break of logic between the last sentence and the discussion before. The last sentence should be moved to a more appropriate place.

Line 244: the notion "high frequency error" needs some explanation. The corresponding "low frequency error" should then be a clear notion as well. About which errors are you talking here?

Line 277: Again, "the error is uncorrelated". To what?

Lines 283-289: Try to rewrite these sentences. For instance: "It is not clear whether the assumptions ... when we reduce the sampling noise. We repeated .... in order to test this."

Caption of figure 2, last sentence: Is it possible to describe the meaning of the grey bar without "alpha channel". Probably many users don't know what this is, and to my opinion it is not really needed.