The slow sweep time would hamper testing of, say, an amplifier that was showing intermittent instability out of band. A conventional analyser can sweep a 2.5GHz span in a few milliseconds. I think my HP8566B does it in 0.02 seconds meaning that instability and rapid changes/peaking in the noise that herald the onset of an unwanted amplifier oscillation can be spotted really easily as the display looks really fluid.
Yes, the 8566B can be pretty fast, especially with option 002, considering that it has over 20GHz bandwidth it better has to be. However, and it's been a while since I last used one, but if I remember right sweeping a 2.6GHz span with the shortest sweep interval (IIRC 20ms) will result in uncalibrated results.
Also, don't forget that a VSA doesn't sweep, i.e. during each acquisition it measures the full selected span, not just a moving window like a swept SA. And while a swept SA requires a short sweep time to capture intermittent signals, this isn't equally true for a VSA.
A lot of reasonably modern analysers that get classed as swept types can also behave as VSA with suitable SW running on a PC. i.e. they offer the best of both worlds with a VSA BW of 10MHz or more. These have user controls for video bandwidth as and when required and can also provide VSA functionality as and when required with the use of suitable software (from Agilent).
As I said, the E7495 doesn't have video bandwidth controls because these only exist on swept spectrum SAs but not on VSAs. VSAs do the same with FFT Averaging, and that setting can be changed on the E7495.
And yes, of course you can do FFT on a separate PC connected to an old SA like the 8566B, and if you can live with having a huge box of 50kg and with a volume of somewhat half a m^3, sucking roughly 650W (and even in Standby it's still around 40W!) while producing a lot of noise, plus a PC, and if you can live with a 30+yr old instrument that is fragile and needs regular adjustment to stay within specs, and where most parts have been obsolete 10 years ago already, then I'd say go for it. These old SAs are great.
However, for those that don't have the space, and want something that is compact, fast and reliable for a limited bandwidth, then devices like the E7495 are much more attractive than the outlook of an old antique on live support. Which is the reason why the market for SAs not only consists of large high end analyzers with 20+GHzs bandwidth, and why there are quite a few portable units. It's not a one size fits all.
Generally dismissing portable units like the E7495 because they don't perform as well as a huge high end lab SA is a bit silly, really. Not every task requires the bandwidth or the absolute performance of a calibrated and fully working 8566B, like not everyone needs a 5+GHz scope. Often enough the lower performance of such portable units is more than good enough. The E7495 doesn't perform badly, it has quite a bit of phase noise, but on the other side frequency stability is really good (even better than for the 8566B!), as are the values for DANL, and sensitivity isn't that bad, either. It's certainly more than just a toy.
The other thing is that the E7495 is not just a SA. It's also a RF generator, a Power Meter, a simple 2-port Network Analyzer (amplitude only though, not phase), and a cable fault localisator. To get the same functionality with an old swept SA like the 85xx Series you have to stack more expensive boxes (Tracking Generator, VNA bridge, Power Meter) on top of the already large SA, which takes up more space, more money, more energy, and introduces more points of failure into already fragile equipment. And still it's utterly worthless once you need to take the setup with you out in the field. No problem with the E7495.
Furthermore, the E7495 is a rugged unit, it's designed and build for use in adverse weather condiitions and with the expectation that it will see at least a certain amount of physical abuse, after which it still has to work 100%. This and considering that these devices are relatively young (I think they started selling the A variant in 2007 and the B in 2009) their reliability and remaining life expectancy is worlds higher than even for the best 85xxA/B Series SA you can find today, and very likely even exceeds many of the current new benchtop devices. It also does that without requiring regular re-adjustment, aside from the internal time base (which just needs GPS lock and is adjusted by the press of a button).
Does the E7495A offer a (1Hz RBW) noise marker function? This is something I would consider essential on a modern analyser for correct analysis of noise signals.
No, it doesn't (but that is something that could be easily done on a PC). But it offers other functionality instead, like a Spectrogram (which shows the frequency band in the time domain with Histogram), measurement of Occupied Bandwidth (OBW) (with user-settable limit), Adjacent Channel Power or Interference Analysis. Especially OBW and Interference Analysis can be pretty useful.
AFAIK the Rigol analysers offer this as a standard feature although I don't know how accurate their system is or if it offers the same noise marker performance/features of the classic lab analysers from the big names.
Quite frankly, from what I've seen from others' and my own Rigol kit, I wouldn't assume that it performs anywhere near the same feature in a big brand SA.