Last week we listened to a meticulously restored Otari at 15ips and universally preferred simple transfers to tape of SACD downloads to the originals. The system was almost totally bespoke valve based amplification/speakers. The general opinion (this was not blind) was that the low frequency HD issues with tape played a role.
This MX50 has the original electronics, but with the opamps swapped out and the appropriate caps replaced with "audiophile" electrolytics and polyprops. I've also built up extensive sections of the schematics in LTspice and performed simulations. At one point I also replaced the entire record/replay chain with tube circuitry (not my own design), but that was abandoned.
The mechanics are pristine, with a new idler wheel etc. and proper lubrication. The W&F has been measured, and is well within spec at (although not as good as my friend's Studer A810, but the difference is inaudible). The record/read back paths have been properly calibrated- I have the appropriate MRL tapes- and the response measured, so that it too is within spec. The head "bump" at 15IPS is less than 1dB as I tweaked the 100Hz setting to trade off 100Hz flatness versus overall LF error.
I can play back tapes- I have a number of 15IPS NAB safety masters taken from a CBC radio station, plus other more recent releases- digitize them, and play back that output versus the tape in a blind AB mode, and as yet no-one has been able to tell the difference- for whatever that's worth.
Playing back the recordings of hires digital versus the digital version in a blind AB test with carefully adjusted volume matching, does indeed seem to sound "better", and the speculation is that the head bump contributes, but that also the increased benign harmonic distortion and the soft limiting of the tape medium may be part of the process.