Hopefully it is not huge (~160 bits), because using a 2 kB salt would indicate that Jim has no clue how computer security works.
As long as the salt is as large as the hashing algorithm's output, it is large enough. In fact, making it any larger will confer no benefit to the number of possible outputs associated with a given IP address, because the hashing algorithm itself only involves a certain number of bits.
Assuming they're still using SHA1, the output will be 160 bits. So long as the salt is at least 160 bits long and is truly random, that leaves 2^160 possible values that any given IP address could hash to. So determining if a given IP produces a certain ID essentially involves running the ID generation algorithm over and over again with different random hashes each time until the desired output is generated, which will on average occur after 50% of the possible values have been tried, which corresponds to 2^159 runs of the algorithm. Another possibility is using some as-yet-unknown preimage attack against SHA1, which will take the equivalent of 2^160 iterations to complete. So brute-forcing is the fastest possible attack.
So given a 160 bit salt, it will take on average 2^159 iterations to crack the salt. So how long will it take to do that?
Currently, the entire bitcoin network performs ~10,000,000 trillion hashes per second. In other words, if you were to use the entire Bitcoin hashing power to attempt to crack Holla Forums's ID salt, the network could perform 10^19 hashes every second. So determining how long it would take for this network to crack Holla Forums's ID salt involves simple division:
2^159 / 10^19 seconds
7.3 x 10^28 seconds
Given 3600 seconds in an hour, 24 hours in a day, and 365 days in a year, that gives a final grand total of:
2.3x10^21 years
Or about 2.3 billion trillion years. In comparison, the universe is ~15 trillion years old, and one estimate for the eventual heat death of the universe indicates that the universe will degenerate into a soup of subatomic particles containing zero usable energy after ~100 trillion years. Cracking a 160 bit ID salt will take about twenty-three million times longer than that.
So to sum it up, assuming secure_trip_salt is only 160 bits, and assuming the entire bitcoin network was pointed towards cracking that salt from a given ID, it would take about twenty three million universes' lifetimes to complete. In other words, the NSA would have to find 23 million more universes and move their cracking operation to each new universe after the old one died until they finally cracked the salt 2.3 billion trillion years after they started their attempt.
This is assuming the NSA will find the correct value after trying 50% of all possible values, which will be true on average. But even if you assume they get massively lucky and find the correct result after much fewer iterations, the time frame is still absurd. Even if you assume they will only have to iterate through 0.1% of all possible salts before they find the right one, it would still take them ~4 million trillion years, still about ~250,000 times more than the time between the current date and the estimated end of the universe.