I was thinking some more about my above post. I realized that, instead of pulling a table of numbers out of thin air, we can derive the curve from geometric-random-walk theoretical model of how prices change over time. Basically you assume the current price incorporates all available information, and moves based on the cumulative effect (sum) of a bunch of small random events. Which means the price at time t will be drawn from a Gaussian distribution with mean equal to the current price and standard deviation proportional to sqrt(t). But this happens in log-space since investments compound, so it's actually log(price) that's Gaussian with standard deviation proportional to sqrt(t). (This is a widely used mathematical model of prices.)
The reason we require 1x collateral is we're trying to confine the probability of a black swan [1] at the expiration time T to some bound p_swan. I'm pretty sure if you believe the geometric-random-walk model, the log of the margin requirement should actually be proportional to sqrt(T). Setting the 360-day requirement to be equal to the original table gives us these numbers:
duration | min_rat
-----------|----------
14 days | 0.31x
30 days | 0.49x
60 days | 0.76x
90 days | 1.00x
120 days | 1.23x
180 days | 1.67x
270 days | 2.32x
360 days | 3.00x
[2**math.sqrt(t / 90.0)-1 for t in [14,30,60,90,120,180,270,360]] # python code to generate above numbers
A 3.0x collateral requirement for a 1-year short implies a 90-day time horizon for 1.0x collateral. But mathematical models are not reality, and allowing shorts with less than 1x collateral may be controversial in terms of adding black swan risk to the system. So I suggest using this formula / table, but forcing the minimum collateral to be at least 1x. The new table is:
duration | min_collateral | order_priority
-----------|-------------------|-----------------------
14 days | 1.00x | collateral_rat / 0.31
30 days | 1.00x | collateral_rat / 0.49
60 days | 1.00x | collateral_rat / 0.76
90 days | 1.00x | collateral_rat / 1.00
120 days | 1.23x | collateral_rat / 1.23
180 days | 1.67x | collateral_rat / 1.67
270 days | 2.32x | collateral_rat / 2.32
360 days | 3.00x | collateral_rat / 3.00
where of course collateral_rat is shorter_collateral / (price * quantity) and min_collateral is the minimum value of collateral_rat required for the short to execute.
[1] An adverse price movement that wipes out the collateral and puts the short "underwater," i.e. the BitUSD that must be destroyed to cover the short is more valuable than the collateral that would be unlocked.