Ok, but we could make a test.

1) Launch a fresh browser and open teams and slack - check total system usage, then close

2) Launch Teams and Slack as electron apps - check total system usage, then close

3) Check the system usage with everything closed.

Probably the closest to a proper test we'll get.

Also, the debate about "unused ram is wasted ram" falls very short when it's storing numbers vs caching documents that are in frequent use.

Your choice of using 64-bit double-precision floating point numbers for things that could otherwise be a byte (or a BIT in some cases).. though I know JS runtimes jump through hoops to avoid this in the modern day, those hoops also cause needless computations and overhead.

When every object is heap allocated with all the cruft that it entales, when it has no value types or structs, when it needs GC headroom and runtime metadata (for JIT) and when strings are immutable requiring more allocations when "mutating"... and also while the ecosystem builds itself around stuffing as much code as possible into the browser... well... that doesn't provide a lot of value for the user..

Preloading the next page I might click and making sure the files I most access on my disk are in memory already: ARE.

Your argument falls short because memory is never wasted if chat programs didn't gobble it all up, it would go to something that the user perceives as higher performance even if it was not used by any application.