I still claim it's 558ns accurate.
You can claim it all you like, but that doesn't make it true. It's approximately 558ns. An accurate figure, determined by a high precision scope might record 555, 559. I totally agree with you that it doesn't matter in system since the circuitry will all be running at whatever speed the clock gives, but the claim it is
accurate to 558ns is not true at all. You have to quote a margin of error for any
real measurement.
All Amiga audio interrupts, cia interrupts, copper lists run at the same speed (NTSC).
They run in the same number of hardware
cycles. As the hardware cycle durations are subject to the timing variances of the oscillator, from system to system they do not run at the same frequency and therefore not a the same
speed.
Again, this is deliberate pedantry, but it is
true that amiga chipset hardware does not run at exactly the same speed everywhere. It only runs with the same cycle pattern everywhere.
For example, some years ago, Redrumloa was experimenting with underclocking the main oscillator on the motherboard as he observied it was giving a speed up as based on common amiga benchmarking programs where the native chipset wasn't being used for video etc.
I was intrigued by his suggestion and a bit sceptical too, so I wrote some software to try and verify his observation. It turned out that all of the original benchmark programs were using EClock to time operations. So, as the system was downclocked, the benchmarks went up entirely as a result. The benchmark code was still running at the same speed but the clock timing it was simply running slower. I was able to test this assertion by using the PowerPC to measure a delay generated using the timer.device and observe that the actual delay generated was getting longer in tandem with the lowering clock speed.
Cycle for cycle, Reds underclocked system was still running the same pattern as any other, but it certainly would not have been accurate to 556 ns.