I don't see how transactions made in 2016 (2014 even?) are part of malicious FUD, they exist and are verifiable. You could assume malice, yet their hypothesis

I don't see how transactions made in 2016 (2014 even?) are part of malicious FUD, they exist and are verifiable. You could assume malice, yet their hypothesis is somewhat testable: if it is uninitialised memory you should have private keys based on word-size shifts due to alignments. 2^64 hashes is somewhat doable. Truncation should be easier to test (in case it's null-terminated).