If I have a soft synth (Omnisphere in my case) playing a basic sine wave, if I play the same note, same velocity, same length, bar after bar, should it be identical for each note?   That is what I was expecting anyway, but when bouncing a track to audio I'm getting different looking wave files for each note, not by much but still not identical.   I thought a basic sine wave should really be exactly the same each time.   Is this the expected behaviour?

Depending on how internal clocks are synced to oscillators, LFOs and the transport, it is very possible to get variations in even the same MIDI performance. I am not familiar with Omnisphere, so not sure what sync options are available, but what you are seeing is a big reason patch points and aux tracks (to record those real-time) was pushed. Some performances cannot be replicated even recording the MIDI because of internal sync.

To add to what mettelus said:
 
There are generally 2 approaches to this in digital synth oscs:
 
1. The osc starts from zero (or whatever fixed point in its cycle) every time you play a note.
 
2. "free running" - this is done to be like analog oscs where the osc is continually running in the background and potentially every note starts at a different point in the cycle.
 
This also can apply to LFO's, and there may also be other random (or seemingly random) differences built into the synth, often to make it "more analog like".
 
 
I don't know about Omni.

Additionally, unless you are using an oscilloscope plugin to examine the waveform, you will generally see the samples taken at different points in the cycle based on the difference between the sampling rate and the frequency of the oscillator.

ULTRABRA
should it be identical for each note?
when bouncing a track to audio I'm getting different looking wave files for each note, not by much but still not identical.

Not seeing what the waveforms look like it's hard to guess what may be going on.  However, if the per-note differences are more than just the phase (position at which the waveform goes from negative to positive), then something is making the sine wave into something other than a sine wave.  It could be the oscillator (maybe applying some kind of "analog-ization" processing) or something further down the line (compression, distortion, who knows?).  A sine wave is a pretty basic building block, and if that waveform is malformed, it's probably not the oscillator doing a bad job, so some other factor must be involved.

Read the Omni manual from around page 400 onward. It talks about the various controls that determine where in the waveform the OSC will trigger from when a note is pressed. It seems it can be set either way. Usually in most quality synths it can be set to do either.