Add redundancy for backup purposes
I'm using the disc spanning feature for backup purposes. I like it because I can burn the files as they are without repacking them. For long-term archiving this is the best method because:
1. Repacking files relies on availabiliy and compatibility of the software used
2. When loosing some sectors of data (might happen for any media), the repacked data might be completely lost when a part is missing
The feature I'd like to have is to fill the last (usually not full) disc with parity data for data recovery. Optionally to even burn additional disc(s) just with parity data. The data should be PAR2-Compatible. This way, you can loose even a whole disc out of a set and you still don't loose data.
Alexander Stohr commented
you can not reliably recover your data with a cross-section additional data disc.
a disc that is dis-integrating has a high chance to become unreadable in a whole. this means the amount of bit errors extends any recoverable level by far.
other chances are that a disc really gets broken or deeply scratched (yes, even the trays can do that when insertion was imperfect or the tray closes too early.)
i assume you will burn all this big data on the same brand and series of bare media. this means all discs are expected to suffer the same aging effects thus damaging the data and the CRC more or less equally in case of any time dependent problems.
if you want a higher safety then burn multiple identical sets on multiple different drives onto multiple different sets of medias from different meida vendors - but each set should be homogenous in itself. a good initial number of sets might be 3 (meaning a first set can be detected beeing fully lost whilst you still have a real chance to reconstruct your data even from the two other sets, assuming that one of them has partial problems.) - if your data represents even higher values you might want to do even more backup copys over time, either as an updated replacement or as an identical "old" copy. - please be aware that same-technology storage can not save you from technology dependent failures like humidity creeping into the metal/color layers or inductivity created damage (like in a metal melting pot coils with high current can "fry" the disc in a contact less fashion where it is most vulnerable - its sort of an EMP danger.) other dangers are heat like from a building fire. - thus use different technology (magnetic tapes, magnetic discs, solid state if found to be reliable, micro films can work for some smaller set of data, and there is the classic paper form for smaller amount of data! - look at NASA, they found out they are replicating their big tower of data permanently just because it is "aging" and additionally because of the used (tape) reading devices are out of production for a long time!)
finally - dont store all sets of the copies side to side. this means you will loose anything at all at the same time in case of fire, water, theft, earth quakes, accident or sabotage. even a high priced cabinet will not ensure more than say 45 minutes (or 1:30) of a chance to still rescue the contents. - for really valueable data take more than 100 miles distance. ideally use different continents for the storage of the sets!
you proposed something that will barely address the typically known dangers. for my understanding this does not solve much - but it rather complicates dealing with the data with an impression for me that the method wont add much of a chance to increase the long term data availability. please do the math - and for my understanding keep it simple. adding more CRCs to an already CRC protected system might not contribute at all in the worst case. (again, do the math on it for a well funded proposal).
My proposal would be to add an independent layer in your treatment. Having e.g. 5 full copies of your data gives you a high chance to read those data low level and comparing the sets to each other bit-by-bit for changes. that is much more of a chance for long term data persistence than a "cheap" but hard to calculate extra CRC disc with a limited recovery ability. - a last hint: dis-integration often works spot alike meaning that lost sectors are "erased" fully and thus have vanished as a whole. a CRC that would be able to compensate for this will have the same size as the lost sector. thus you are well advised to just do more full sized copies of your set.