BladeRF or LimeSDR for running osmo-bts-trx

Hello Osmocom community, (Cc'ing Eric per instructions from Harald) I am looking to acquire an SDR device for the purpose of running osmo-bts-trx. My initial goal is to get to the bottom of some mysteries observed in OS#6579; because it is CSD, getting a non-virtual test setup requires an SDR-based BTS rather than the otherwise wonderful sysmoBTS. And of course there will likely be other fixes and improvements that can be made in osmo-bts-trx. :) Harald recommended that I get a BladeRF 2.0 xA4 for this purpose, plus a Leo Bodnar GPSDO. However, before I commit to buying that hw with an aggregate cost of around $800, I wanted to get some community input to make sure it is the right choice for the job at hand. (There is also the possibility that Sysmocom will sponsor this hw for me, which is wonderful - but still, I don't want to waste money even if it's someone else's!) Harald told me that Eric has worked on BladeRF, and instructed me to ask both Eric and the wider community. My basic question is: how stable and reliable is this BladeRF 2.0 in the BTS application? I presume I will need to run osmo-trx-bladerf, but I find it alarming that there is absolutely no documentation for this variant: the wiki and the user manual cover osmo-trx-uhd and osmo-trx-lms, but not one mention of BladeRF. It is my understanding that BladeRF always runs on 38.4 MHz clock; if the external clock frequency is some other, the PLL inside BladeRF will discipline the local 38.4 MHz VCXO to "follow" the external reference, but the SDR will still run at 38.4 MHz. Thus I see no way to avoid fractional resampling, i.e., configuring the external GPSDO to put out 26 MHz or 52 MHz won't do anything to eliminate that fractional resampling need. Hence the question: does this fractional resampling constitute a real problem, or is it a sufficiently minor defect such that one can live with it in a lab environment? Will there be any gotchas to be aware of, such as MS struggling to connect to the test network? Finally, how does this BladeRF compare to various LimeSDR options? It seems like LimeSDR has always been the popular choice in the "cheap" department, whereas for BladeRF I see not one mention anywhere as I already noted. How do these two options compare? I recall hearing that while LimeSDR was once very popular, people were also having problems with it. Is BladeRF now better in terms of stability, or not really? And if BladeRF really is a better choice than LimeSDR for the purpose of running an SDR BTS on something cheaper than Ettus hw, then how did we get into the peculiar situation of having documentation that covers LimeSDR but not BladeRF? Please help this SDR-ignorant gal figure out which hardware I should buy. Whether I cover the cost on my own or take up Harald's offer of cost reimbursement, either way I need to be sure I buy the right thing... Oh, one more detail: whichever USB-interfaced SDR I get, I plan on using a Raspberry Pi 5 to drive it, running stock Raspbian. Does anyone have any experience with using RPi devices to drive SDRs for use as BTS? Is RPi5 powerful enough to run all of the computationally demanding parts, including that fractional resampling? Is stock Raspbian good enough in terms of sw dependencies? TIA, Mychaela