I would like read ideas about possible EMI/RFI or other electrical issues.
The problem is that EMI/RFI does not cause such minor effects at all.
I know this sounds hostile and unreasonable to you, but here's the thing: we know how these devices work, intimately. What you are describing is like claiming that you can feel the color of the surface you're walking on: not the material or paint type used, but the color itself. If you are human, you do not have cone cells in the soles of your feet actually connected to your central nervous system. It is therefore almost certain that if you really believed it to be so, it was a
correlation: instead of sensing the actual color, your senses had managed to collect
correlated information, letting your human mind connect the
correlation to
causation. Which is very typical of us humans.
Electromagnetic interference and radio frequency interference can affect a computer, sure. However, the way computers are constructed means that either such effects are corrected without any overhead (as in ECC memory, reading data off an SSD or HDD or a CD/DVD/BluRay disc), the error is ignored/overlooked (say, changing just one bit that does not affect the result enough for noticeable effects), or the computer crashes or locks up due to the error.
The processors do not "redo" work when it notices an error. Except for network communications, error correction is done in hardware (and the hardware does whatever it does in the same time regardless of whether there were errors corrected or not). In network communications, the information sent and received is in packets, with each packet containing a checksum. If that checksum mismatches, the packet is thrown away, and the sender has to re-send it.
As an example, consider ECC RAM. It works at a specific rate, each read and write access taking a specific amount of time. This time is listed on the chips or the memory module. The motherboard contains hardware that
during each access, verifies the contents match the extra error correction bits; and if not, "fixes" the data. (This is the difference between
checksums and
error correction codes: the former are designed to indicate integrity, the latter are designed to be useful to fix integrity issues.) If the data cannot be fixed, a Machine Check Exception occurs, and in past Windows versions, this causes a bluescreen. In no case is there a measurable slowdown regardless of whether an error occurred, a correction was attempted, or not: the ECC works at the same rate in all situations.
(Perhaps CDs would have been a better example, because they often have errors –– scratches and whatnot. If the ECC slowed the player down, then even minor scratches corrected by the player (so that no glitch is left in the audio data), the player would slow down and the corrected errors be audible! That would make ECC useless in a CD player. Yet, CD players were a major step forward in ECC use. The only case you can actually hear a glitch from a CD is when the player cannot correct the errors anymore. Because it is a simple device, it often loses its place in the spiral, and starts repeating the same thing. Computers are more complex, and instead of stuttering like that, they just crash.)
I am still assuming there is actually something wrong in your computer setup. The timer latencies you showed in post #21 definitely look horribly wrong.
To
prove it is a "software problem" –– meaning something not electrical, but something in the OS, motherboard, interaction between motherboard and processor, EFI BIOS, or even malware or virus somehow re-infecting your machine –– I would recommend you do a fresh reinstall of your machine, and before installing any games or anything, redo the kernel timer latency test you did in post #21. If indeed a fresh reinstall makes the problem go away for a short while, and it is a software issue related to kernel timers (because they're the one thing in this thread that could cause the issues you're seeing), you'll see the timer latencies be much,
MUCH smaller. And, if/when the issues come back, the timer latencies also grow to the huge values you showed in post #21.
I know that is quite a bit of work (or, rather, time spent for just to confirm a possible cause); but, it is the most likely actionable (as in, "this we can do something with") symptom thus far. If indeed the kernel timer latencies correlate with the problems you observe, and do not occur in a freshly reinstalled system, we at least have a clue we can work from. (While I personally cannot tell you how to fix that kind of an issue, I can help with how and where to report the issue to get a fix.)