I'm interested in a variation of https://oeis.org/A006339 (Least hypotenuse of n distinct Pythagorean triangles).
Basically, if A006339 is $u(n)$, I want $v(n) := \min_{n \leq i} u(i)$
In other words, $v(n)$ is increasing and for any $k \leq n$, there are less than $v(n)$ triples with hypothenuse $k$.
Are there tables for that sequence? How can I compute it?
A possible way would be to have lower bounds for A006339 and compute it until the lower bound is big enough. I'm not even sure about how I can compute A006339. There is an easy enumeration algorithm for Pythagorean triples due to a result by Hall (1970) and Roberts (1977) (https://stackoverflow.com/a/8263898/5133167), and it allows us to stop the search as any multiplication by $U$, $A$ or $D$ will increase the value of $c$. Is there a more performant way?

Worked it out. Taking the consecutive (1 mod 4) primes $$ 5, 13, 17, 29, 37, 41, 53,...$$ we calculate the best exponent for each such prime $p$ in terms of the real positive exponent $\delta.$ As I said, this is Ramanujan's approach, used in his Superior Highly Composite Numbers. I learned it from Nicolas
Given real $\delta > 0,$ we demand the exponent of $p$ to be $$ k=\left\lfloor \frac{1}{p^\delta - 1} \right\rfloor $$ Meanwhile, with prime $p$ and desired exponent $k,$ we use $$ \delta = \frac{\log (k+1) - \log k}{\log p} $$ This defines a sequence of numbers that rapidly increase $r(n) .$ Indeed, $n$ is guaranteed to have more representations as the sum of two squares than any smaller number has.
As was Ramanujan's case, there may be new champions of $r(n)$ that lie between two of the numbers defined above. In that case, there is a procedure due to Guy Robin, I think in his dissertation (supervised by J.-L. Nicolas). I don't know how complicated his "benefit method" is, but I programmed such things (operations research) as an undergraduate, and recall that a fair amount of work was involved. The constraints include: the prime exponents are non-increasing and always non-negative.
Alright, Nicolas gives complete details of the "benefit" method in ACTA 1988 . The approach of Robin is to combine numerous tools. This is the article where he displays the smallest number with more than $10^{1000}$ divisors.
I can write a program to show the first few such numbers...
$$ \bigcirc \bigcirc \bigcirc \bigcirc\bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc\bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc\bigcirc \bigcirc \bigcirc \bigcirc \bigcirc $$
We get an ordered list of consecutive useful $\delta's$ by solving for each prime and each exponent up to $5.$ To make sure it works properly, let the $\delta$ used be in between two consecutive reals indicated in the list below. An alternative is to read the sorted file, ignore the actual value of $\delta$ on that line, and multiply the number $n$ by the prime on that line. If done properly, factoring the resulting $n$ at each step should show the increased exponent for that prime $p \; . \; \; $ $$ \bigcirc \bigcirc \bigcirc \bigcirc\bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc\bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc \bigcirc\bigcirc \bigcirc \bigcirc \bigcirc \bigcirc $$