Parity check reads all the data drives, does the parity calculation, reads the parity drives and compares. If it's correct, it moves on to the next sector. If it's wrong, parity errors are incremented, and depending on the write corrections setting, it either leaves the error as is, or writes the calculated value back to the parity disks.
Normally you don't want writes to happen unless you know WHY it's wrong to begin with. For example, after a forced shutdown where writes to the data drives were occurring, parity errors are somewhat expected because the data writes are committed first and parity may not have had a chance to complete.
Under typical operation you should NEVER have parity errors, so finding out why ( bad cable, bad RAM, etc, etc) and fixing the root cause before committing the writes is the prudent thing to do.
So, normal routine parity checks, NON-correcting, if parity errors are found, figure out a plausible explanation and correct the issue, including any file system corruption, then do a parity check with write corrections, typically followed up by a non-correcting check to be sure whatever was wrong is now fixed.
Zero parity errors is the only acceptable outcome. Any errors means a failed drive will be reconstructed with the wrong bits.