Oh stacking noise is a huge issue with sampling.
I sometimes think that every (media) composers formative education (or even later when they're in the world trying to make a living from this artform) should include learning how to make a sample instrument. Start with something basic - and do a project like Christian Henson's MVP Piano. Indeed - if anyone here hasn't done it before and has a piano, go ahead and contribute to that project NOW!
Anyone here who uses Nuendo can build really simple sample instruments right in their timeline. It's awesome. Then when you want a little more flexibility with things, learn a little about Kontakt. (The Logic sampler is quite good too!)
Anyway - my point is - this will quickly show up how sampling is a very different artform to recording. Even though it involves recording.
As Jeremy pointed out, voice stacking quickly adds up.
For fun, go grab a detailed piano kontakt library, and play with the pedal down for a while. Maybe some ostinato / textural phrase. It is easy to reach 300PLUS voices playing back at once. Each of those voices can be made of more than one mic (if using a "mix" palette it will often be 6 or 8 even for just a piano. So its quite easy to get a situation with just a single instrument to have 1000's of individual signals summing to make the one sound. Noise - no matter how low - will play a part.
My conversations with various developers over the years have shown me how the attention to noise floor is paramount. Even when they're using character equipment.
Some devs go for super super quiet mics, preamps etc. Even they can need post production (you'd be surprised the amount of work done on samples before they hit a final product in kontakt!) . Some devs do like tape / using tube gear etc. Used well, this isn't as quiet as solid state gear, but it is still very quiet - and it definitely changes the character of the samples. But perhaps it is not to taste for composers who need things to be super super quiet.
Add to that the need for media composers to present palettes of sound that could never be reproduced by a real recording without riding faders. I'm talking about super soft pianos, strings, orchestral instruments. Its an artificial dynamic created by gaining up the samples significantly.
End of last year I recorded a quintet + piano for a score in a super quiet concert hall + studio in Glasgow. Great gear, an awesome engineer and producer.
I was using loads of super soft playing / textural orchestration. Not too many mics. Spots, Tree and Surrounds. Never more than a single overdub. And yeah, the engineer hit me up a few days after delivering the files with new files he had additionally denoised as he felt it really needed it.
And during the mix - oh did we have noise to deal with. Funnily enough, I got both used to it - especially after getting the temp mix from the dubbing stage for the rest of the film to mix "around'... and I totally embraced the noise in the end - even for the super quiet emotional pieces. For me it really worked - but I was on the edge at a time.
I also played back the premixes (prior to live recording) of the noisiest cues - and they used what others would consider noisy samples - and the final result was definitely quieter than the recording.
(We *chose*) on the recording to use ribbons on the piano for the sound - knowing the noise there would be the noisiest bit! But it was a good wakeup to the immense work the sample devs put into getting us the tools we do have.
Most of the time I can tell noise issues in samples before buying them. Just listening to quiet demos etc. And reading reviews. I've never personally come across a time when it has stopped me from using the sound... or just changing up an idea in response to me not liking a result etc. Creative constraints and all that.
I choose to celebrate the tools. Yeah, it can provide frustration at times. I understand that. But also, samples are bloody AMAZING! I listen to what I did 25 years ago knowing what sampling tools were around then / libs etc - and to hear what is possible - just WOW!
So I'm just not sure I subscribe to the idea that because there is inherent noise involved in a particular sampling technique that a company shouldn't offer it to market. It isn't in everyones taste and its a square peg at times for your round hole. But there's plenty of square holes I need filling at times...
And I appreciate the character that comes from some equipment choices and especially room choices. They're all choices. The more I think about it, the more it feels similar to the room / reverb sampling arguments that have been going on for years. Both sides have their place. Sometimes we just need to realise our tools are not all one size fits all or even most.
And celebrate the tools for what they are.