Since we're discussing them: I've been using my Cobra again since getting a second A1200 (my first 1200 got a Blizzard 1260 upgrade in the late '90s and the Cobra then went into storage). It's a 28 MHz model with a 33 MHz 68882 and 8 MB Fast RAM also installed (I have previously used 16 MB and 64 MB SIMMs with it), no SCSI.
It's been solid and stable overall. I've experimented with overclocking the CPU, which requires desoldering the oscillator, and if you use bare socket pins to hold the replacement you must be careful to insulate the metal shell of the oscillator lest it short the 5 V supply! It mostly runs fine at 33 MHz but hangs after a Ctrl-Amiga-Amiga keyboard reset (while troubleshooting I noticed dual pulses on the _KB_RESET line but this is apparently normal). Reinstating a 28 MHz oscillator seems to have fixed that; the stock crystal is 28.63636 MHz (NTSC) but I had a spare PAL 28.37516 MHz one and the machine is PAL otherwise. I guess it's picky about timings. Notably, the firmware can be updated (see
http://amiga.resource.cx/exp/cobra) with variants for 28 MHz, 40 MHz, and 50 MHz CPU clock speeds and 70 or 80 ns RAM timing, so I think the timing in the glue logic is fairly critical. The 68030 is actually a 20 MHz part but presumably is run at 28 MHz to keep things synchronous with the Amiga chipset.
You can run the FPU at a different clock speed from the CPU without issues (there's a socket for a second oscillator and a jumper for this). The MapROM function (hardware jumper plus software) is effective.
Mine did develop a fault where the real-time clock was failing to keep time and the battery was draining rapidly. This turned out to be due to a bent pin on one of the ICs that was shorting against the battery holder! It might have got bent during installation one time.
I see IDE-CF speeds of around 1.6 MB/s in SysInfo but haven't tried any IDE speed patches on this machine.
