I've been scratching my head as to what you are experiencing. I'll first explain a bit about how the "experimental" oscillators work and then share some more specific thoughts and comments about getting to the bottom of your experience with the feature.
How the Oscillators Work
The TXo sampling rate is at 15,625 Hz across its four oscillators. This means that the top frequency that each can produce is 7,812 Hz.
The oscillators themselves do no oversample. This means that sine waves will produce pure tones up to the top reproducible frequency (the nyquist rate). Other waveforms (triangle, saw, and square) will alias. This means that they will create ghost overtones or digital distortion that is present in the output.
As you may have noted above, sine waves should not alias when performing a simple oscillation. Rapid changes to their frequency or CV min/max could potentially introduce aliasing into the signal. A tiny amount of oscillator frequency slew can help abate this a bit, btw, if you are going to be programmatically changing the frequency rapidly.
The oscillators in the TXo run from what are called lookup tables. The tables in the TXo are currently made up of 512 values - although that size is planned to grow in a future firmware update. The size of the lookup table is directly related to the quality of the signal (bigger being better).
The TXo implements a technique called linear interpolation to improve the quality of the oscillator regardless of table size. It estimates values in-between the points outlined in the table. This technique is active when oscillators are used without additional features being active.
Once you turn on any extended features for that oscillator channel (envelopes, frequency slew, rectification, ...), interpolation is disabled in order to provide the appropriate processor resources to the voice. This means that you will see some additional distortion if using an oscillator in an extended mode, even with the sine oscillator.
We've got more room for a larger lookup table after some recent memory optimization, which will improve quality when interpolation is disabled. A future firmware update will most likely increase their size.
Your Experience with Oscillators
I'm trying to figure out what is going on with your demonstration. Could tell us a little bit more about your oscilliscope and method for altering the frequency. What does your Teletype script look like? What does your oscilliscope look like showing the output of an analog oscillator doing similar things (with similar waveforms)?
Congrats again on the build!! Very glad that you got them up and running.