That is not a bug nor wrong. If any, the original firmware is wrong.
A signal has rise time and even if it is very fast, the scope itself is not due to its filters. The new firmware plots the signal in a standard x, x+1 way, so the last pixel of the square low period is on x, and the first pixel of the square high period is on x+1, hence the need for the vertical line to have what you call a kink in it.
To make this smooth dithering could be applied, but that makes the code more complex and slower. To make it straight the code needs to scan the samples to see if there is an almost instant change in the signal and then make it a straight vertical line. And this would be falsifying what you see, which is what the original firmware does a lot.
In the new firmware what you see is much closer to the real world signal. Sampling is based on making discrete steps in time, which leads to what you see. Sample A is behind in time on sample B, so no line can be straight in the vertical direction. Due to noise it ain't always straight in the horizontal direction either.
Even using a sinc filter would not improve on this issue.
My advice is to study the theory of signal behavior and sampling and use a proper scope for comparisons instead of making comparisons between the original firmware and the new firmware.
The FNIRSI 1013D is and stays a cheap scope and will never become as good as even the cheapest Rigol or Siglent scopes. Even the Hantek DSO2000 series is better than this FNIRSI will ever be, no matter what amount of software or FPGA configuration development you throw at it. The hardware is seriously limiting.