This paper uses an OSSE to examine the effect of lagged SWIR detector response on Sentinel-5 retrievals of CH4, CO2, and CO. The work appears to have been carefully done and is clearly explained. But it is technical in nature, narrowly specific to Sentinel-5, not particularly original, and in the end the effect turns out to cause relatively small errors. It uses an OSSE formulation originally developed for MicroCarb that the MicroCarb team apparently did not bother to publish (it is cited as personal communication). I would expect the work described here to be in a Sentinel-5 ATBD, or as part of a broader Sentinel-5 publication, not as a stand-alone peer-reviewed publication. At the same time, I welcome information on Sentinel-5 in the peer-reviewed literature.  I didn't know how to express this in the "For final publication..." buttons and the only reason I clicked on "major revisions" is for the editors to decide whether this paper passes the bar for AMT. It is marginal in my opinion. But if the editors decide that it passes the bar then only minor revisions are needed.

Specific comments:

1. Lines 13-14, and elsewhere: The abstract and conclusions refer to CH4 errors as those from the full-physics retrieval and just mention “proxy” without reference to what it means, but a strength of Sentinel-5 is precisely that one can use the proxy method for CH4 (unlike TROPOMI). The proxy product will receive far more use than the full-physics product for CH4 and it should be presented as the default product for CH4.
2. Lines 13-14, and conclusions: Giving maximum errors in abstract/conclusions is misleading without the error statistics, which would show that the errors are in general considerably smaller than these maxima. In fact, what I get from this work is that the detector lag will not significantly affect my use of the proxy product for CH4 – I have other errors to worry about.
3. Figure 1 is not very helpful. It would be more helpful without unfamiliar abbreviations and with a more substantial caption that describes the overall OSSE strategy.
4. Line 120: I didn’t understand reff = 4. Is it missing a unit? But um would be too large and nm too small?
5. Line 120: I didn’t understand “Surface albedo spectral dependence is modeled as wavelength-independent.” Does this mean that the albedo spectra are just scaled up or down depending on surface? That wouldn’t make sense because the wavelength dependence of the albedo should vary with surface type.
6. Line 199 and elsewhere. I don’t understand why the detector lag would affect the CH4 proxy retrieval at all, since the albedo cancels between CO2 and CH4, except for some smearing which is maybe what the “one pixel” refers to? But that effect would be independent of change in albedo and it would not translate into a systematic bias.
7. Line 205: I don’t understand the nits unit. Aren’t the % errors reported for column mixing ratio?
8. Line 260 and elsewhere: I think that the authors’ recommendation to exclude fractional water scenes is too radical and would result in unnecessary loss of information for important locations such as coastal cities. The errors reported are not that large, and I certainly would not want to filter out proxy data on that basis.

See supplement.

This paper presents a simulation study to quantify the effect of detector persistence on trace gas retrievals from Sentinel 5. The study looked both at radiance scenes with a surface albedo step as well as three different true radiance scenes. The overall bias observed was significant and larger than the mission target parameters for single pixels around surface albedo transitions; however, the paper shows that these can be filtered out by applying radiance change masks. Overall, the paper is well written and clearly describes the work. I only have a few minor comments:

P2 l58: Can you add a reference for the MicroCarb and CO2M requirements?

P3 l79: "where CNES model applies a temporal sampling of the incoming signal". This was unclear. Would it be more appropriate to call it a "temporal smoothing" and then also reference the appendix where it is described in more detail?

P4 l116: Was HITRAN08 used for all of the line parameters? Why? What profile was assumed?

P6 l165: I'm not sure what the '(numerical range)' indicates

P6 l175: 'extend' -> extent

Fig7: font very small