FYI, both look like control instability.
The first, some sort of chaotic failure mode where the current regulation, or whatever it's controlling on, goes not just unstable but bonkers and makes trash. Current is still roughly following the line voltage, so power factor is okay, but the error is bounded rather than asymptotic. That is, it bounces around some range above and below the setpoint, rather than approaching it smoothly as a stable and linear controller would.
The new one looks far better behaved, but is subject to its programmed limits -- namely that it must shut down in brownout and overvoltage conditions. So it starts up, puts some current into the line, line voltage overshoots, it scares itself, shuts off, oh would ya look at that line voltage is nominal, starts up again, and so on. Probably this continues until either a load switches at the same time it does (canceling out the bump), or until enough load pulls down the line voltage generally to where it can start without hiccuping.
AFAIK, the behavior is required by rules, though whether the particular limit here is overly tight or something, I don't know. It is Bad Design, from the standpoint of, it's not only a wildly nonlinear scheme (it's switching on and off suddenly over not much change in voltage), but it's a negative dynamic resistance (higher voltage --> lower current) so is predisposed to oscillate. Perhaps they could've decided upon a more stable set of rules, but, it is what we have now.
Possibly there is a startup speed or power limit option for it that you could adjust, or maybe the voltage threshold, or calibration even (but, the latter would mess up your (onboard) metering, so beware of that).
Possibly, you could add some rotating mass at your end of the line, to act as a sort of rotating capacitor to smooth out fluctuations -- this could be as simple as a large flywheel on an induction motor with power factor correction capacitors, I think, but is more often a synchronous machine; and perhaps a rotary phase converter would perform a similar job, if you happen to also need one for other purposes (good way to run higher powered machine tools from single phase
). This wouldn't be a cheap solution, though.
You could also run everything through a power bank, if you have one (or are interested in so investing). This would be charged gently by solar and/or mains, as needed, or return to the grid likewise. Perhaps throughout the day as demand and grid pricing varies (you'd combine this with a time-varying pricing model from your utility), or just slowly as the solar/consumption balance varies from day to day. The lower load on the grid would also reduce how much and how often it changes, mitigating your impedance problem.
I'm finding that my tech support journey isn't necessarily that much better. It's just too far away from what they typically deal with. Somewhere in the depths of the organization's head-office hides a grey-beard who would shit himself if he saw this trace. But coming via first level support in another country - it's hard to get past the gate-keepers.
FYI, as an engineer myself, I have 0.01 reason (not exactly zero, but low!) to believe there is such a person -- in many projects, I might be that very person, or I might be cleaning up after someone hacking their way through a project. Especially from China, designs are highly evolved -- copied from western designs, reproduced and modified, parts substituted with local variants, and lightly tested before release to production and sales. Only shallow understanding is needed in the process; and even those doing original from-scratch work, can get by on account of low wages and cheap parts/proto costs, with evolving rather than synthesizing their designs.
Compounding that, our most frequent interaction in the west, with such designs from the east, is through budget-priced wares -- which, well, what more do I have to say, y'know?
And just to make that clear, this isn't at all unique to China, it's just more prominent there, given their problematic educational standards/methods, and the above bias. There are also plenty who are highly capable -- they have an IC industry for example, many have tried to do that throughout history and most have failed, it is truly one of the most advanced multidisciplinary industries to pursue -- but top performers are in short supply and, I assume, eagerly sought out by companies in most dire need. This is all true of western countries, just differing by degree: there might be more and better here, but the average engineer still amounts to a technician with a leveled-up degree.
Tim