You mean it's not enough that you really have to over produce on solar by a long margin and waste a bunch of energy during the day and hope you never fully deplete it at night
although maybe a running cost in terms of power might be possible if charging cost is in there (although I'm pretty convinced it is included) maybe as much as 10% and then a trickle drain all the time a day is 25000 ticks and I'd think 1% per day is a good number (and we'll ignore the compound interest of what I'm about to suggest) so 1%/25000= 0.00004% per tick so just deplete accumulators by that amount of their current capacity, it's negligible unless you're living on the edge. Adds realism for you and won't upset me because it's barely a nerf.
Added complexity joined to a slight buff and realism of accumulators maybe make it so it's maximum output rather than a fixed number (300kW) is a % of charge so actually the battery never fully depletes and over the night each battery becomes less and less effective, on the other hand charging is the same a % of the empty space is filled per tick and so battery management becomes a thing. You could have them all plugged in and discharging at the same rate and at the beginning of the night you might have more capacity than you need but because the batteries output isn't constant over the night you'll be using more and more of the potential output, and brown outs in order to keep that uniform output you'd need to provide enough output at the beginning of the night and slowly connect full batteries to maintain power. But that'd leave all your accumulators at a partial charge in the morning and so as a uniform lot might take longer to recharge as a result
currently 5MJ of stored energy takes approximately 17s to fully charge/discharge at the maximum rate of 300kW
translated into MW ticks that's about 300 MWt of charge at a floa 0.1% full capacity per tick
Instead we go for that same 5MJ/300MWt and 0.2% of current power. Initially you get twice the power output of the current setup, I like that, you say it's OP we agree to disagree, by 5.75s rather than 8.5we're at half charge and the accumulators ability to output is now level with current values.
It would take until a full 13.3s for the total power outputs to match, it's at this point the old accumulator has output more power than the new one for the first time, and it's just going to keep giving power but the new accumulator is outputting a fraction of the power of the old one and is losing ground rapidly, as much as a third less power. Already and dwindling. It will keep giving out power for much longer keeping the whole factory running at a slower speed while the other one shuts down production all together but you'd definitely prefer to be running at full capacity and you might even have enough energy to do that in the accumulators
The wiki suggests that 25 accumulators will provide 1MW of energy all night, 2500 ticks (41.6s) my own maths says 9 (well 8.333 but we'll go with 9)
Trial A 9 all on all the night no staggered power on. Brown outs start to appear at 36.6s (11 necessary to have no power loss)
Trial B 9 staggered switch on 2 switch on at first needing 1MW the initial 0.2% isn't enough, the third switches on at 1.7s but it's only a trickle, fourth turns on at a staggering 6.7 seconds, fifth at 11.7 seconds and so on until the 9th at 31.6s meaning brown outs at the same 36.6s and the same 11 for no power loss but now some will be much easier to charge than others. And getting to that peak charge be the primary focus of power management leaving the most depleted batteries until last, or charging them first and trying to keep the charging power usage even over the day rather than needing a big boost at the beginning and trailing off. etc etc..
If these were turned into curves rather than linear effects the complexity would be multiplied and new solutions would abound
Or leave it as it is because if you change efficiency you don't actually effect the balance/how OP something is you just change what ratios are plopped on the map and built if people can over produce and ignore the complexities.