Let's start with the RAW. This is the sensor data as well as being read-only image data. You can return to this file forever as if the image file had just been offloaded from the camera today, even years after you've discard the original camera body. This is the true digital asset to carefully maintain and back-up, etc. RAW files are likely to always be available for editing by ongoing software titles. Once a camera's RAW format is supported by any software title, it will likely be supported always by ongoing versions of that title.

The exported JPEG from that edited RAW, yes, is a 'permanent' file, and the file is available for ongoing editing by an even larger population of software titles, being a 'universal' image format.

The XMP sidecards are the edit instructions to convert the RAW into the JPEG. They are not the edit instructions of the JPEG. If you lose the RAW, the XMP has no purpose as those instructions are dependent on the input source file, the RAW.

The JPEG has your edit decisions baked-in like any JPEG, including the bit-depth lowered to 8-bit and probably having the colorspace reduced from the largest ProPhotoRGB that is used internally by LR. It's a 'better' JPEG than would have come directly from the camera, but the JPEG is an end-result / output format when compared to the original RAW. All the rich original data has been compressed into the smaller (output) JPEG format. The JPEG is definitely not an analog to a 'film negative'.

The absolutely first step is to cull, there is NEVER a reason to keep / edit 200 image, and certaily not 500, of any topic EVER.

If you shoot in RAW, you need a tool that renders an image at the pixel-level details so you can kick inferiors in just a 1- or 2-seconds review. That rendering needs to be just as fast as your 1-second kick decision. Instantaneous. FastRawViewer is one such tool.

So, 200 images, 200 seconds, less than 5-minutes for cull 1. Now, run a second review / cull of the images, where you're looking only at potential keepers against other potential keepers. You should be down to 50%, to possily 80%, of that original offload. Your very best from that original 200 is now maybe 40, 60, 100, if you were having a really good day of very unique and distinct topics.

Now, your workflow and duration and level of effort are dependent on your choice of tools, and your level of technical skill with those tools. Personally, I use almost exclusively Adobe Lightroom Classic. I have Develop Presets that I trigger after the import of my culled images. I use LR's sync technology to sync the editing of one similar image against others that were imported with the same editing needs, things like exposure adjustment, WB, movement of the blacks, whites, highlights and shadows. Things like sharpening and noise processing were defaulted to the camera chracteristics and ISO-level of the images via the Develop User Presets.

All through my LR editing, I continuously seek to kick inferior images. If I imported 40 images, I'm still trying to drive down and isolate only the best, whether that's a final 10, 20 or 30. It won't be 40.

Time-wise, that I try to avoid any complicated edits, except where needed. I'll level and crop to improve the compositions, but removing content is a last resort, only if I don't have another image that avoids this complex and time-consuming work. That's why I shoot lots of versions and have many to choose from to find the single best that was the whole reason I was capturing that subject / stretch of images.

Although I remember having maybe 1000 images from the Chicago Marathon a few weeks ago, I have only 46 images kept in LR and 566 images in the recycle bin from where they were offloaded. I may have deleted already those images from the cull that would fill-out to my 1000+ memory. My LR timestamps from import through export of those 46 keepers is a 7-hour timeframe, but I watching TV and doing some other things too during the edit effort.

Chg Canon has usually sage advice and observations. Time to process is highly variable and I cite myself as a prime outlier for PP time. I work as a product photographer for an online auction company. We have literally thousands of objects to photograph. Think of your house, car and everything in the house and that is what I will shoot. Production is rapid and high throughput is required. It is not uncommon for me to shoot between 500 to 800 photos a day. Thank God for assistants to set my shooting table for the shots. Obviously high-end items get more time, but a box full of kitchen utensils can be done in less than a minute and that includes close-ups. The lighting needs to be dialed-in and I typically shoot in aperture priority and set it at f-11 to get a reasonable depth of field and still retain overall sharpness.

To get to the OP question. I track my hours for PP and I can run between 100 to 150 photos per hour. That includes all operations from uploading from camera to computer, importing to LR, editing, exporting from LR and then uploading to the internet bidding site. I do most of my editing in LR with the occasional operation in PS. I can do a lot of fast tweaks and have developed a number of presets that really speed up the process. Of course, using LR "synchronize" function allows me to apply my fixes to any number of photos.

The key to operating at such high volumes day after day is in my estimation a matter of exceptional lighting that works adequately well for most if not all shots. The second key is using a full frame camera with a 24-105mm zoom lens. I rarely have to switch to a wide-angle lens and that is often for large Persian rugs or a large piece of furniture in a cramped or narrow room, and then in most of my real estate shoots. Even with my fast throughput I still get complaints from management of the amount of time in PP. I'd say 99 percent of my shots make it to the website for public viewing. Given the constraints, I think I do fairly well.