How practically? With real numbers? The computational effort required to find a suitable collision in this scenario is drastically reduced from the theoretical
2^256. A well-resourced attacker (*cough* Lazarus *cough*), targeting the lowest-entropy available signed hash(es) could feasibly obtain the necessary ~2^32 to ~2^40 uniquely 'mined' collision attempts in milliseconds to seconds. Want some perspective? Let's take a top of the line Bitcoin miner (yes, I know they're ASICs) - Antminer SP19 (or w/e it's called). Let's say the latest model out can generate 140 TH/s (per unit). That's 140 trillion hashes per second btw (or 1.4 * 10^14). If some well resourced group repurposed an FPGA (or stack of them) to 'mine' candidate hack transactions and was able to generate the equivalent hash rate of just 3 bitcoin miners, they would bust through all 2^32 or 2^40 candidates in MICROseconds and MILLIseconds, respectively (if that's what our search space is constrained to).
Even if we up the search space to 2^60 (which is an exponentially exponential leap from 2^40), the hash rate of 3 Bitcoin miners is cracking that in under an hour (45 minutes approx.). Even at an upper bound of 2^70 (which would put us close to impractical / impossible territory), an equivalent hashrate of JUST THOSE 3 miners would find a collision in about 32 days.
Lazarus Group has made out with >$6 billion in this space over the last 3-4 years alone. You gotta be smoking crack if you don't think these dudes are probably operating in the nicest state-of-the-art Bruce Wayan style cave you can think of. North Korea has a great relationship w China.
Why wouldn't they be able to contract a little chip foundry out there, get some FPGAs rolled out for pennies on the dollar (when considering the take-home)? Well? Why wouldn't they employ it for this use?
They just made out with $1.5 billion. If these guys spent every second of their entire lives up to that point planning for that hack, you'd still have to consider that to be a profoundly wise investment of time (from a risk/reward ratio; not cheerleading antisocial behavior).