Exposure Shift in Timelapse After Syncing Settings in Lightroom

I was thinking of answering with a reply similar to richardplondon’s, and the only reason I didn’t post that earlier is because you’ve done a lot of time lapse photography, so I thought maybe you might already know about the issues with image-adaptive adjustments. But now that your testing seems to confirm that it does have to do with adaptive adjustments such as Shadows/Highlights, I agree that a solution is to do as much tone adjustment as possible using the Tone Curve, and if that isn’t enough, then add Shadows/Highlights and other Basic panel adjustments, but only as little as needed.

If this is the first time you’ve come across this issue, watching the video below is highly recommended, where the developer of LRTimelapse software goes into great detail about how to avoid exposure flickering induced by Lightroom Classic adaptive adjustments. The last part of the video, starting about 13 minutes in, is a revealing analysis of exactly which adjustments cause the most and least flickering when synced across many images. (From his article Stop Lightroom from adding Contrast Flicker to your Timelapses)

Also, note in the video that he demonstrates how some unwanted shifts may be made worse depending on whether a camera or Adobe raw profile is applied! This was new info to me.

Looking at the edited samples I do wonder if this could be arising from the nature and strength of adjustments used. Some Basic adjustments - shadows / highlights and clarity in particular - are more image-contrast-adaptive: meaning that for each image they are applied to, even when the slider is moved to the same value, a fresh assessment will be made for each image on how it should apply (a pixel of the identical starting value may receive a different adjustment within the two images, depending on what other tonalities it accidentally occurs in proximity to). The reality is that physical shutters are not perfectly repeatable from one operation to the next, so the absolute camera exposure may vary slightly even if everything is locked down on Manual. Also shot noise and to a lesser degree read noise is inherently variable shot to shot.

So very heavy use of a few strongly image-adaptive adjustments will tend to magnify any accidental variations between the original shots. More moderate use of these adjustments and the strategic combination of different adjustments will tend to do so much less, perhaps below the threshold of noticeability, all else being equal.

In my experience it's far better to (for example) combine a moderate general Contrast slider adjustment with moderate Shadows / Highlight adjustments, rather than (for example) trying to achieve everything just from setting extreme values for the Shadows / Highlights sliders.

Also, Tone Curve is not image-adaptive at all AFAIK, so this can act more as a neutral and constant tonal 'overlay'. Note highlights recovery for example, must be addressed properly in Basic panel adjustments: Tone Curve cannot optimise such things absolutely. It can only relatively darken or lighten whatever the Basic adjustments are producing.

I think the only effective way to narrow down the issue is to share the original two raws and the develop settings.  Select the first raw and do Metadata > Save Metadata To File. Then upload both raws and the .xmp sidecar for the first one (which contains the Develop settings) to Dropbox, Google Drive, or similar and include the sharing link here.

"DirectX: NVIDIA GeForce GTX 1650 (31.0.15.4609)"

That driver is about seven months old. It's probably not causing your issue, but it's always good LR hygiene to be on the latest driver. Go to Nvidia.com and install the latest Studio driver.