> permanent read errors
I wonder if you read the tape in, say, 51 times and the tape might read a bit a certain way more or less than half the time, if it would fix such errors. E.g. if we have five reads of the same byte on tape:
01010110
01010111
01010110
01001111
01010111
The byte is most likely 01010111.
There could be software to calculate most probable bits over a certain number of reads.