We had dozens of 2013 MBP with discrete GPU mux where I worked. I bought one afterwards and used it until the M1 laptops came out, for pretty much everything you can use a laptop for. Never had or saw this problem in any of them, FWIW!

As did my 2015 Thinkpad - same problem, both an integrated and an external GPU, and I was constantly fighting with the drivers to get it to just make use of the external GPU. I'm plugged into a wall whenever the laptop isn't suspended/shut down for transport, stupid machine, stop making Autodesk and my games crash because you want to boost your battery lifetime numbers!

They had a special Lenovo driver that would occasionally become overriden by Windows updates but could be reinstalled manually, I dual-booted Debian though and getting the system to work properly under that was a nightmare. There were a couple years when I simply gave up, I got it to work with the iGPU and I wasn't running anything more graphics intensive than a browser so I simply left the discrete GPU idle while running Linux.

Incredibly frustrating.

I think all these manufacturers are desperate to get their published specs for battery life estimates to double-digit hours that can't be reached while running the discrete GPU at full speed all day. Heck, they can't be reached while running the CPU at full speed all day, you're not going to run a 35W processor and a 55W graphics card and a 20W display (10W when you arbitrarily reduce the max brightness when on battery power) all day.

You've got like 90 watt-hours available in the battery, at 100% usage on everything the real capacity is gone in under an hour...which is unacceptable. So Asus and Apple and Lenovo and everyone else have to come up with some hack to turn it off whenever it's on so that the spec sheet says you can get 8, 10, 12 hours of runtime.

Yeah, that's not shocking. Linux uses the Intel ACPI infra, whereas Windows uses Microsoft's. It's as good as they can do, but it's not going to be possible to perfectly replicate Windows, let alone improve such clearly buggy firmware, at least generally.

It is possible on Linux to override some of the firmware (most notably the DSDT, e.g https://wiki.archlinux.org/title/DSDT because so much hardware is broken). So, if you can make or get a fixed version, you should be good. A wholesale replacement of all the ACPI assets, though, seems unlikely. I could well be wrong, though.

Anyway, in this case, I suspect the poster was advocating for Macs.

Or the Installer.app, at least from 2015 to whenever they finally switched to image-based updates.

A percentage of users were "unlucky" to hit a bug where Installer.app would end up in infinite loop trying to unpack a pkg file when updating OSX. My personal record is minor update that ultimately took a week.