They way PCM and value-hold DACs were explained to me in the past was that PCM at 44.1 is a series of discrete (digital) samples of a recorded continuous (analogue) sound. Based on the math of the famous Nyquist–Shannon sampling theorem, as long as you sample 2x of the highest band limited frequency, you will be able to perfectly reconstruct the original analogue waveform, from a series of discrete values. Now as far as max frequency of your DAC goes, let it be 44.1 * 16 = 705.6 or 48 * 16 = 768, that would be the max resolution of the DAC, so that the input PCM file of said frequency on a value-hold DAC has the most detail possible. Why would you want to upsample from original input of 44.1? So that the jumps between samples (if you use value-hold aka zero order-hold) method (no interpolation/approximation between samples) so that you do not have jumps over less resolution (44.1) but more resolution (705). In essence, it is putting in new discrete values, so that jumps between samples are not as high.

Nyquist–Shannon sampling theorem has one requirement, that is quite difficult to achieve in practice - it requires "a sinc function, where these sinc functions are summed into a continuous function. A mathematically equivalent method is to convolve one sinc function with a series of Dirac delta pulses, weighted by the sample values. Neither method is numerically practical. Instead, some type of approximation of the sinc functions, finite in length, is used. The imperfections attributable to the approximation are known as interpolation error." I believe this is what M Scaler tries to do, it uses better interpolation, they call it number of taps, as infinite is not possible, more taps you have, better the result in practice.

You asked for a laymen's explanation, I did my best to provide it, I hope I did not butcher it completely. Not sure how much of this is audible, but my understanding is that if you had infinite sinc (or diract pulses) you would be able to reconstruct perfectly. Again, worth repeating is that I tried both HQPlayer and MScaler with a blind ABX testing, and I did not like what it did to my music.