Where is comes to actual equipment recommendations:
Many base station analysers have spectrum analysis and network analysis features. I got one from JDSU / Viavi which does both. These typically don't break the bank. The JD745B/JD785B models could be interesting but make sure to understand the available options versus what is installed on the units.
I also got a JD785A, which one can get on eBay for ~$2k with the SOL cal kit. It do have okay RF performance on datasheet, in particular for SA dynamic range (2/3 (TOI-DANL) in 1 Hz RBW: >104 dB) and phase noise (typ. –112 dBc/Hz at 1G, 100k offset), and it's modern, portable and hackable. Data saving, remote control and transferring is a piece of cake as there is LAN and USB, and one can be creative as there is a modern linux system running on armv7 inside serving SSH, HTTP and FTP.
Also 7x5A and 7x5B seems to be identical except the useless RFoCPRI function, unless you want to play with RRUs.
However, one should understand their limitations before getting one (or E7495A/B or other Agilent handhelds that are not FieldFox), not only the limitations visible on the datasheet like the dynamic range of the VNA mode is only 80 dB (which is
not better than the budget Siglent 3.2 GHz VNA, the higher frequency Siglent is 20 dB better IIRC), and also hidden caveats like
1. AM/FM demodulation is not continuous but only limited to a certain record length.
2. Marker readout is too slow to use.
3. The CW generator is noisy and have poor harmonic rejection.
4. Somehow the video triggering for gated sweep doesn't seem to work, it always stays in free running.
5. Bugs exist like recalling VNA settings with calibration can led to garbled reading. (the latest and likely the last firmware revision)
6. The VNA does not measure phase and magnitude simultaneously (at least I couldn't get it to), but need two measurements to get the complex data.
7. Forget about using the cellular analysis functions to look at generic complex modulated signals, it will likely require a significant amount of hacking.
8. The VNA only uses a very very basic calibration model, assuming the cal kit is ideal.
If I had 2x of the available budget, I should have got myself Siglent SSA3032X-R instead of this. But hey, for only $2k one can get a bundle of a 8G SA and 6G VNA, they should work for your transceiver development despite their limitations. The handheld form factor make it not only ideal for field work, and also in the lab much less head scratching moments realizing the cable is not long enough.