I have a 12 v relay driving an large 220 volt AC relay . Across the contact
of the driver relay i placed one RC snubber circut (27NF with 100 R
resisitor in series) to help with some spikes that were influencing the low
voltage driver circuits.
The driver circuit is able to detect mains zero crossing and fire the
driver relay at an angle i choose .
From what i read the best point to switch off the power relay is at zero
crossing . I did that and i show a large spike up to 1 KV at the relay
contact followed by a decaying 500hz waveform to 0 volts . After some
experimentation the best point came exactly when switching off at the peak
of the mains voltage .At this point there is smooth decaying waveform to 0
volt after 5 periods of 500 HZ but no overshoot. The relay presents no
arcing. If i remove the snubber and make the experiment the best place to
switch is zero crossing but i also see large SHARP spikes up to 500 Volts
Peak.

WTF are you talking about, the arc across the driver relay contacts or
the arc across the "large 220 volt AC relay?" It gets reall aggrvating
when you don't make yourself clear...

My question is
The switching with snubber must be made at zero crossing or at the peak of
an ac voltage waveform ?
What is the behaviour of the circuit ?.
As i understand any large spikes can harm the X2 capacitor i'm using so
what is the best operating practise ?.


Your whole idea is dumb anyway, you're working with relays and pull-in
times that are a substantial fraction of line voltage period and vary
with age and operating conditions. Snubbers are spedc'd around worst
case and not timing. They are only an approximation, the inductance of
the "large 220 volt AC relay" is non-linear and may differ by 10:1
between pulled in and out. And some relays require an arc for contact
longevity. You let the relay manufacturer take care of the switching and
just work out a better 12V drive circuit immune to the dV/dt current
from that HV ringing, usually means lower impedance drive.