This is a partial answer. I will add to it as I work out the details.
Exiftool (https://exiftool.org/) can be used to extract lots of interesting metadata tidbits from images.
Starting at my image directory:
exiftool -r -ImageSize -Model -DateTimeOriginal -ImageNumber .
Image Size : 299x400
Camera Model Name : iPhone 4
Date Created : 2015:05:17
Image Size : 267x400
Date Created : 2015:04:29 12:58:16
Note that all data isn't present for all images. One of the apps on my phone, "Night Camera" puts very little metadata in the image. Note also that sometimes I just get a date and not a full date/time. Turns out there are 12 time stamps. I changed the command above to -DateTimeOriginal as it seems to be best supported.
If it seems that I'm a bit nuts about dates, there's method to my madness: When I import images, they get renamed yyyy-mm-dd_hh:mm:ss so with some changing of : to - and space to _ I can get the actual original filename.
The -r in the example command will go down my image tree recursively. Save this to a file.
Check it to see that every file has a meaningful date. I may decide to pull further info.
This is also my time to rationalize my folder structure. Over the 10 years I've had my web page, 'things jest grew'
My thinking right now is that I will export a version of each image as a 3000 x 2000 pixel jpeg. That's too large for now, but hey, Bill Gates said no one would need more that 640K ram in a PC.
Using ImageMagic I will then create a set of images from each large image reducing pixel dimensions by a factor of 1.414 (square root of 2) with a sharpening stage between. See http://www.controlledvocabulary.com/imagedatabases/downsampling.html for details about this.
ImageMagic is a suite of command line tools for image manipulation. The syntax is extensive, somewhat arcane, and it's generally not a tool for use on a single image. But if you need to work on a thousand or so images it saves a lot of time.
The same script I write to use ImageMagic will also write srcset snippets.
Right now, I write my webpages in Markdown, so an image looks like this:
Caption goes here
This will get replaced with
[% INCLUDE srcset/Shelterbelt-3.sst]
Caption goes here
All that is manual work, but because the srcset file tree is separate, there will be continuity as I change each image. If a user has the older cached image, he's no worse off than before.
Intent is to go after the pages most frequently used by mobile first.