For PCB Rogowski coils, this paper is pretty good:
https://www.mdpi.com/2079-9292/12/5/1099/pdfThis example design attains a mutual inductance in the 20 nH ballpark. 15 A RMS at 50 Hz has a peak dI/dt of 15 * sqrt(2) * 2pi * 50 Hz = 6.6 kA/s, giving an output voltage of 20 nH * 6.6 kA/s = 130 microvolts peak, or just under 100 microvolts RMS. This low output level, combined with the increased sensitivity of the coil at higher frequency (rogowski coils have sensitivity linearly proportional to frequency) makes accurate measurement of mains frequency hard.
I wonder if a CT could be practically made with some finite and well controlled magnetizing inductance, but with a magnetic core to push the mutual inductance into a more interesting range for low frequency measurement, making it practical to implement for 50 Hz measurements in the form of a planar transformer. It would be constructed like a CT, but used like a Rogowski coil. The main challenge is that sensitivity would depend directly on magnetizing inductance, which depends strongly on the air gap, and also on core permeability, both of which are hard to do with fine tolerances and good temperature stability.
What are the exact requirements? Putting some numbers to this can make it simpler to find a suitable solution. You mention that you don't find ICs rated above 100 V, but there should be tons of options here, open loop hall effect sensors with internal (ACS712) or external (MXL91208) current conductors, closed loop hall (LEM LKSR-series) and flux-gate (LEM CKSR-series), GMR-based sensors (ACS70331), isolated amplifiers for current shunt measurements (AMC1200), even ones with integrated power isolation (AMC3306).
If I was to design the cheapest and most compact multi-channel current sensing system for a PDU, I would strongly consider resistive shunts. If you place them before the relays, then they can all have the input feed in common. This way, they can all be read by an ADC referenced to a single potential, making it cheap and easy to have a large number of channels. A multichannel ADC would sit at the mains potential and read all of the shunt voltages, and the interface can be done over isolated SPI for instance.