Using a multiple to upsample is an easier job and hard to mess up. Using a non-multiple helps with jitter but is harder to implement.
As for up/over sampling before the DAC, it matters whether either side takes care of intersample overs, the Benchmark models for instance does take this into account and lowers the level before oversampling.
Interpolation does “destroy” the original data while synchronous upsampling preserves the original data. It matters how good the interpolation is. Going with your currency analogy, it would be like if I had a $100 USD but someone took it and gave me $100 in Bahamian currency, which averages a 1:1 conversion with USD, but sometimes it’s not, it has been worth 0.98:1 and sometimes 1.02:1.
Stereophile’s measurements of DACs shows how much aliasing occurs by using a high frequency test tone, if you look at the AudioQuest DragonFly Red’s measurements for instance, the aliasing image at 25kHz is below -100dB.
I can’t comment on phase though as I haven’t seen that tested before, but we are very forgiving of phase errors (not talking phase mismatch) in the treble that I doubt any decent DAC will have an issue.