>> I think I know what happened here: swapping uses _all_ of the swap
>> partition, [...].  So when you put swap at offset 0, you ended up
>> swapping all over your disk label (and possibly boot blocks,
>> depending).

*blush*  I spoke too quickly.  It turns out there's code in
vm/vm_swap.c specifically designed to avoid using the first few blocks
of a pack:

		if (vsbase == 0) {
			/*
			 * First of all chunks... initialize the swapmap.
			 * Don't use the first cluster of the device
			 * in case it starts with a label or boot block.
			 */
			rminit(swapmap, blk - ctod(CLSIZE),
			    vsbase + ctod(CLSIZE), "swap", nswapmap);
		} else if (dvbase == 0) {
			/*
			 * Don't use the first cluster of the device
			 * in case it starts with a label or boot block.
			 */
			rmfree(swapmap, blk - ctod(CLSIZE),
			    vsbase + ctod(CLSIZE));
		} else
			rmfree(swapmap, blk, vsbase);

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu