In this case, it's a complex story. NSF has a collection of astronomical observatories (optical and radio).
Some of them have decades-old equipment, and keeping everything operating will eventually preclude building anything new. So, NSF has been trying to offload some older observatories to universities, or consortia of universities - Arecibo is one such.
NSF's budget has has been anemic for the last 15+ years so projects get built then the maintenance budget evaporates in favor of some new program dictated by Congressional pork barreling. It wouldn't be a problem if the NSF budget grew at a steady pace, but the budget has only grown when one party has complete control of government and has shrunk otherwise (I'll leave it as an exercise to the reader to figure out which party is which).
I'm on my way to look up DKIST, but after SOHO, it's hard to imagine needing ground based solar observatories. Of course, SOHO will eventually come to end of mission, so we definitely need other observatories. It's just the SOHO imagery is stellar.
You’re a little out of date! SoHO is largely shut down, has been for a decade now. Its tent pole instruments were replaced by SDO in 2010. (https://sdo.gsfc.nasa.gov/mission/)
Both of these are NASA, not NSF, observatories. They were (SoHO) and are (SDO) amazing resources that have advanced the whole area of heliophysics.
DKIST (NSF) has a 4m aperture. Unbelievable for a solar telescope, and of course, impossible to fly under reasonable NASA heliophysics budgets.
Some of them have decades-old equipment, and keeping everything operating will eventually preclude building anything new. So, NSF has been trying to offload some older observatories to universities, or consortia of universities - Arecibo is one such.
For instance, some older solar observatories were taken offline to build DKIST (https://en.wikipedia.org/wiki/Daniel_K._Inouye_Solar_Telesco...), which is an amazing instrument, just now coming online.
Of course, there is a lot of strife and anger caused by reallocation of resources like this.