Australia's largest power plant has 2.9GW of inertial generation assuming all generators are running at 100%. As in the small battery substation alone comes close to the countries largest power station. I'm not sure where the idea that lithium ion can't dump power quickly comes from. They are absolutely phenomenal at it. Australia's building dozens of these substations too since they are so cheap and reduce overall power costs. It's a win from all points of view.
Large spinning masses can provide several seconds of inertia. For 2GW of traditional turbine, you would have between 10-20 gigawatt-seconds of energy that is instantly available at any moment to resist RoCoF.
The whole point with actual inertia is that you get a large multiple of your maximum capacity without any redundant parts or added system complexity.
Keeping around 10x+ more semiconductors than you need to cover a tiny fraction of operational scenarios is difficult economics.
A semiconductor device cannot be overloaded like a spinning generator or transmission infrastructure can. You cannot trade temperature and maintenance schedule for capacity in the same way. Semiconductors have far more brittle operating parameters.
More inverters in parallel will achieve the same end goal - fast frequency response.
(And "reactive power" could be good too but not absolutely necessary to understand at first...
Basically I'm dubious. I'm sure there are grids somewhere that have misprovisioned their inverter capacity, but I don't buy that battery facilities are inherently unable to buffer spikes. Is there a cite I can read?