3 run validation is not described anywhere in the GMP’s, and is not statistically significant. That being said, 3 run validation is the traditional means/methods to do a Process Validation (the FDA calls this PPQ in the 2011 guideline). Ultimately the FDA is promoting a more robust validation using design of experiment, statistical evaluation, and ongoing monitoring (see recent guidelines and ICH’s)
But following the old school of validation - I was taught (by Mark Witcher, presenter at various ISPE conference) that PV can exclude a failed lot, if there is an assignable cause that shows the process was not followed, and is therefore not representative of your procedure. I think it this case, you make a claim that the process, when followed as prescribed, shows repeatable results (3 lot validation). Thus, you can exclude a lot from a process validation, as long as there is an assignable cause, which you described.
Of course, the question now is - why wasn’t the process followed? Was your batch record unclear, were your people untrained, were there proper checks and balances, etc. So even though your process for manufacturing can be shown to be validated, your quality systems might not be. So you’ve gotten through the process, but your business might have other issues besides the manufacturing process.
This is not a clear agreement or understanding in the in the industry on this topic. Often QA will assume it is not worth the dialogue/risk, and that it is safer and more conservative to just make the fix, and do a new 3 lot validation.
Personally, I think that a well defended justification can actually show more process understanding, rather than just a “fix it, re-do it, and re-test it” mentality.
Like Boomer_Chemist mentioned. Validation should not be a one-time activity, it should be coupled with ongoing understanding, process control, and process understanding.