P: Auto Tone of a batch of photos gives different results than if you do them one at a time

I very much sppport the above suggestion to delegate the decision to the user rather then just silently assume everyone is willing to sacrifice fine tuning in a favor of performance. Even if applying Auto in the best way will take hours for a thousand of photos I Would just initiate the process in the evening and it would be done next morning. As long as I don’t have to spend hours staring at monitor and clicking Auto for every image to have *the best* auto conversion - I’m fine and I don’t care how hard my computer will work to produce a better adjustments for my photos.

Missing from the recipes for 2, 3, and 4: Do Quick Develop > Reset before purging the cache.

Great to see such a response to my first post.  This appears to be a "Feature" no a "Bug" then, but maybe one that Adobe should think about.  Up until now I have used "Auto Tone" as a good start point but always ended up adjusting the best photos further.  So slight differences had no real effect on workflow.  The new AI based version seems like a game changer and will allow far more photos to be just automatically adjusted.  Unfortunately when that "Auto" button is live it is very difficult not to push it!  So if we are looking at a compromise between speed and accuracy, what about a preference setting to allow us to make that decision ourselves?

In Develop I selected a batch of photos on the filmstrip, right clicked on them and selected "Develop Settings, Auto Settings" (I think the option used to be "Auto Tone")

Hmm, the results of applying batch Auto Tone via Quick Develop are non-deterministic and sometimes depend on whether 1:1 previews exist or have been rebuilt a second time!   Most of the differences seem small, but with one photo, there was a difference in exposure of +0.86 versus +0.64 -- whether that's "slight" depends on one's subjective judgment of the photo, of course, but it didn't seem slight to me.

The table below shows the results of applying Auto Tone to seven pseudo-randomly chosen unprocessed raw photos from different cameras.  All the results had a standard preview size of 1440.

Row 1: After doing Reset and Auto Tone in Develop.

Row 2: After doing Preferences > Camera Raw Cache Settings > Purge Cache; exit LR; delete Previews.lrdata; restart LR; Build 1:1 Previews; select 7 photos; Quick Develop > Auto.

Row 3: After steps in row 2, then Discard 1:1 Previews; Purge Cache; Build 1:1 Previews; Quick Develop > Auto.

Row 4: After Purge Cache; exit LR; delete Previews.lrdata; restart LR; Build Standard Previews.

Based on what Simon described, this makes me suspicious that the Purge Cache command may not be fully flushing the raw cache.

Based on your description (the auto tone button is active and clicking it changes the settings slightly), you apply the one photo Auto in the develop module.

This is what happens, when the Auto button is activated, it means Lr has loaded the full resolution negative. So the Auto is computed based on this full resolution negative.

When you apply Auto to a batch of photos, for performance reasons, Lightroom loads a lower resolution negatives cached in the camera raw cache (like Smart Previews) and computes Auto from that. Since the two input are different, the results could differ slightly. 

BTW, the issue exists in the previous version of Lightroom as well. It is just that the new Auto shows more differences. To summarize, it is a performance vs accuracy tradeoff.

...or Quick Develop?

Which method are you using to apply it to a batch: Auto Sync, Sync, Copy/Paste, or a preset?