Single bit flip can definitely cause *any* kind of corruption, it just depends what bit it was. Obviously to cause major damage you need to be somewhat unlucky. I usually have ECC ram on all computers, but on lesser stuff I have had some non-ECC computers. On those, I have seen all modified cache blocks to be written on disk as zeroes, continous random blocks written to wrong places on disk, and number of other incidents during last 20 years of having memory-error linked filesystem corruption on non-ECC computers. Not using ECC has cost me a lot more money in debugging and rescuing messed up servers and workstations than the savings on cheaper non-ECC memory, so I always find hardware with ECC. It does not really cost that much extra. For budget, HP microservers have ECC and 4 disk slots, and they are something like 200€ per unit. Supermicro has neat Atom server motherboards with ECC. Workstation motherboards are a bit more difficult as most consumer stuff, in particular Intel stuff does not have ECC or BIOS ECC stuff has been left out. AMD may have better price/performance ratio in some cases, when ECC is requirement.
Single bit flip can corrupt ZFS, just like any other filesystem. It does not help to have checksumming for blocks, as the corruption can be in metadata, memory pointers, memory data structures, and happen, say, in driver which writes blocks to wrong place, ignores checksums or generates wrong checksums, bit flipped before calculating checksum. The system can write too much or too little data, leading part of data not written at all, or valid data written over with whatever was in the memory. The list of possibilities what can happen on single bit flip is endless. Written data may be correct but it just got written over to some metadata in some wrong place. Raid features will not help here either, as the same corruption may go to all devices. After this, you have corrupt metadata, and likely, if this is not detected, you will continue to get more corruption as long as the system runs and does not stumble on the corruption, which may happen far too late to be able to rescue anything. ZFS or any other filesystem does not and really cannot protect against that. It is designed to detect disks returning wrong data to you, not memory errors.
Not using ECC is way more expensive than using ECC, if you put any value to your working hours, and properly value risks of your own or customer's data being corrupt. ZFS does NOT protect you from most memory errors. It does reasonable job protecting you from disk and disk-memory transfer errors. And it does not have fsck, so if it gets confused, sorry, restore backups. Restoring a large ZFS filesystem takes time, by the way.