For archival storage and instant project access, I use Darktable under Linux, presently with redundant high-capacity (8TB) external drives. Some specific or special shoots or projects may be further exported from here into Cloud storage. Otherwise for a creative editing project, Darktable is used to select and copy data in various ways (select raws, or other fotos intended for edit); or, just one of the 8TB archival drives is used with a my editing machine (described next) to obtain a whole shoot, or select fotos, depending in the project. Projects are imported by Darktable by type, then by date, along with supplememtary keywords, and sometimes EXIF adjusts -- like fixing having forgotten to update the camera's clock when first in Europe, etc. In this manner, I've finally been able to facilitate centralized access for work across 15 years of time, for some 450+ foto shoots. Think of this as the darkroom, and the main negatives storage cabinet.
Then for creative editing, per the above, data is moved to a literal light table: my Google Pixelbook, which is a combined tablet and laptop computer. Touching the image literally for me is much more like working with film. For this processing, I often use the Linux container on ChromeOS as a secondary and transient negatives storage area, where I'll run Raw Therapee to perform detailed camera raw edits and conversions. Then specific creative edit work is done on ChromeOS with Polarr and Snapseed, plus some glamour-specific utilities. Creative product results from here go where they go, online posting, upload to cloud for magazines or print, now and then actual print-making. Then the finished art is often archived back to the "main negatve storage cabinet," but more often Cloud storage, which becomes the "Product Warehouse" storage facility. Finished work is grouped into folders by project and date.
Today, no software from Apple or Adobe ever touches my art work, none whatsoever. All original foto data and project data is ultimately used only transiently, never stored on the edit machine archivially, in case it fails. A broken edit machine is annoying for sure, but replaceable. Lost data is not. The "main negatve storage" is operated under a Triple Modular Redundancy (TMR) policy, to which I'm adding additonal geographically dispersed storage voiumes. I do not believe in RAID for local physical storage, but I do believe in blob storage in the Cloud, which is exponentially more reliable than RAID. Drives are operated only when they are needed, preserving mechanical and magnetic storage life to 《 20000 hours operating time, in order to stay ahead of the MTBF curve. I have found that digital data is nomadic, it must move from time to time, or it will fade away with the sands of time. Approximately every 36 months main storage is upgraded to the next doubling of storage capacity that costs about US $130. This is generallly the knee in the price/capacity curve, and read/write bit rate always improves with new technology. As new work gets added to the collection, improved technology enables keeping full data copy times down to 12-16 hours per volume. I am definitely managing millions of my own fotos, mostly of belly dancers, but enough other work that I accidently discovered that I am also my own stock foto agency. I'm considering adding machine learning adjunct to Darktable, to make finding specific things faster. Otherwise, in terms of the "Product Warehouse," Cloud storage is as reliable as Civilization is. Which does feel a bit more risky these days, and it does take money and faith.