I have guitar effect that has quite small sampling rate(31.250kHz) with oversampling and I wonder whether it is enough to convert signal from other effects, especially distortion, or it is designed only for raw guitar signal.
All other things being equal, higher sampling rates are generally seen as better as they allow higher frequencies to be reconstructed when the signal is converted back to analogue.
Simplistically speaking, the highest frequency that can be reconstructed is half the sampling rate. This is why 44.1 KHz is a frequently-used sampling rate - it allows frequencies up to around 20 KHz - commonly thought of as the limit of human hearing - to be reconstructed, with a bit of room to spare (which may be needed for the anti-aliasing filter).
This is fine (in theory) if all you are going to do is listen to the signal. However, let's imagine that we are going to do further processing - for example, pitch shift the output down. You might want frequency information beyond 20KHz (and therefore a sample rate beyond 44.1 KHz) to allow those ultrasonic frequencies to be audible when they are shifted down.
So, even 44.1Khz can't always be said to be 'good enough' if you are going to do further processing. 31.25 KHz, then, is 'even worse'... but it's not that much worse. You're only losing about 5 semitones right at the top of the human hearing range.
Also, remember that fidelity probably isn't the aim of the game here - guitar signals are often band limited at the top to prevent them sounding unduly harsh. The very slight roll off at the top might even be a good thing.
For most applications, I wouldn't worry too much about it. If it's a good effect otherwise, there's no reason to think this technical limitation is a musical limitation.
If it sounds good, it is good.
If the effect is well made, there will be analog filters prior to the input A/D converters. These analog filters will suppress frequencies everything above about 15.5kHz (1/2 the sampling rate) before the signal gets digitized. This will have the effect of cutting off the highs in the processed signal, but shouldn't be a problem (if the pedal is well designed). Technically (i.e. if you ran it through a spectrum analyzer), the processed signal will be less bright, perhaps even noticeably so to your ears. You might not even notice this if it is a reverb or delay or similar that mixes an unprocessed version of the signal (I'm assuming that the unprocessed path doesn't go through the filters) with the processed (and thus high-cut) signal.
If this is a modulation effect that is attempting to "sound analog" the lower sample rate is (probably) OK -- analog delays tend to roll off the highs anyway, so you might not need as much bandwidth to reproduce the sound.