the flawed rng was in libbitcoin-system, used all over the place the supposed thread safety bug that motivated the change was observed in P2P networking
disclaimer: i'm the wrong person to speak about thread safety, not enough experience
...but, the
pull request doesn't seem to make any changes that should alter threading behavior, anything that does is incredibly subtle (which I appreciate is often the case with thread-safety bugs)
sounds like I ought to check if Armory's use of libbitcoin could be affected. I didn't create any new Armory wallets since this the date (late 2016) of the libbitcoin-system pull request, but possibly others did. Armory previously used Crypto++, which I guess was subject to a little more scrutiny (guessing isn't good enough however, I feel compelled now to check)