Solar thermal also makes more sense where you're directly utilising the heat generated, eg: for industrial processes. If you're just using that heat to spin turbines to make electricity, you're adding a layer of inefficiency.
In contrast, solar thermal runs at extremely high temperatures to begin with, so even in a hot climate, there's a huge temperature difference from which to extract power.
It seems to me that mirrors would be cheaper than photovoltaics, and the power-plant itself is a fixed cost, so most of the cost must come from the motors and electronics needed to angle the mirrors, and the maintenance of that.
I wonder: could you have a system where the mirrors are all fixed in place, and you simply move the target as the sun moves. (Or at least move a smaller set of mirrors/lenses that refocus the light onto a fixed target)
There's also a nice scalability aspect to PV: if you want to expand your plant's output, it's really just a matter of adding more panels as you need them. Where as once you hit the limits of your collector/turbine/etc, you pretty much need to build a whole new plant to expand it further.
One of the world's largest solar farms is in Dubai, obviously an extremely hot region. It contains a mix of PV and thermal technologies, though it sounds like they are leaning towards more PV for future phases:
https://en.wikipedia.org/wiki/Mohammed_bin_Rashid_Al_Maktoum...
This site https://www.scottmadden.com/insight/solar-photovoltaic-plant... estimates $60 per kW-year for maintenance. Half cleaning; most of the rest in inverter replacement. How does that compare with turbine I wonder.
Cleaning is certainly a cost in dry areas where dust can coat the panels. But generally not in temperate climates where you can rely on self-cleaning via regular rainfall.