People generally speak about bit rot in regards to RAID rebuilds, as it is the worst time to encounter a unrecoverable read error.
Issues with rebuilds can generally be boiled down to 3 areas. hardware failure (during a rebuild), user error, and poor maintenance.
I have seen lots of users try to "recover" a RAID volume, and just make things worse for themselves. If you are not sure what to do it is better to speak to the vendor on how to proceed.
Anyway, if you run regular volume checks, then this will ensure that your data and parity data match and a rebuild will go very smoothly. Most MFGs of hardware RAID cards recommend daily or weekly volume checks depending on the size of the array.
In regards to RAID 5/6 parity data is not a copy of the original data, it is a calculated representation of the data, that needs to be extracted to fully represent the data in question. Due to this design, if anything bad happens to this PAR, it will become corrupted as the controller will not be able to extract it.
With RAID 6, you would need the original bit to switch, as well as both parity locations to become corrupted for the controller to not accurately know what data should be in that location.
During a rebuild things are more vulnerable, as only one set of parity data is available, so if the bit doesn't match the PAR for that sector the controller must decide which bit is correct. Most controllers have CRC/hash/integrity checks on the PAR data, so they can tell what is corrupted, the PAR or the original data.
Also this is where maintenance comes into play. If you run regular volume checks, then those 3 locations (or 2 if you are rebuilding) will almost always match.
If the disk controller cannot fix an issue with CRC or other data integrity mechanisms, the RAID controller has it's hand at it. Typically this will resolve your issue. However if not then the error is presented to the file system layer, which there are some good choices for resilient file systems.
Most places with RAID will not stack a resilient FS over the RAID volume, due to how robust RAID6 or even 10 is. However it would work if you wanted to do this.
We runs lots and lots of volume checks on lots of TB's of data. And we almost never see any volume check errors at all, and I have never seen an uncorrectable one on a RAID6 array.
It can be scary when you know a rebuild is going to take 2+ weeks to finish. I am much more worried about other hardware failures (other HDDs or controllers) in that 2 weeks then "bit rot".