Kerbal Space Program

EC consumption with robotic parts has been changed trying to improve balance, but your example crafts show how it wasn't enough.

Indeed there's a serious issue with the EC consumption with robotic parts. When the frog starts jumping, the net balance of EC = 129.95 (constant). When not jumping but the fuel cells are active, the EC balance is = - 30.17 (constant). Meaning all those parts with legs consume 129.95 + 30.17 = 160.12 EC/s. With a maximum EC stored = 2005, it takes 15.4 seconds to totally deplete EC, even with the fuel cells on.
The issue is those robotic parts keep consuming at full power even when don't produce any work. In the jump cycle, work is produced only when the legs are extending and pushing against ground to make the frog move. This happens for a fraction of the cycle, the first 0.5 seconds in how you had it set in KAL. During retraction, legs move in air so the only work done is against drag (and friction inherent to those robotic parts). It requires a lot less torque to retract legs in air (perhaps 5% of the maximum), and no torque at all for those 2.2 seconds while legs don't move; but torque currently with robotic parts is the same in the whole cycle. Less torque means less power required, so the total EC consumed for each jump would be a lot lower if torque changed during a cycle.

Now, if anybody wants detail, how much is the energy from each jump? The largest part goes with the work produced to move the frog up. Frog mass = 11.256 tons, jumps about 3 meters high against gravity at 9.81 m/s^2, so the work produced for just jumping high = 11.256 * 9.81 * 3 = 331.26 KJoules. In KSP 1 EC ~= 4500 Joules, so the above is = 73.61 EC. Note some more energy is spent to retract legs and provide the forward movement, let's estimate total consumption with each jump at 90 EC. Fuel cells produce 30.17 EC/s above whatever consumed while the frog doesn't move, so would be enough to keep the frog moving at a frequency of 1 jump every 3 seconds.