The snubbers should be across each bulb, not the junction box starting at the closest bulb to the main junction.
i don't get why it matters where you put the snubbers.
I'm assuming our hypothesis is that the flashing of LED's is being caused by an 'induced' /capacitively coupled voltage arising from the circuits switched live conductor(s) close proximity to permanently live conductors.
If you consider the equivalent circuit of what's happening, you have the capacitance (and inductance?) of the coupling between the cables (no resistive component unless there's a fault!) , and the capacitance/resistance of the snubber, acting as a potential divider, with the lamps across the snubber side. The resistance of the copper wire is absolutely negligible in relation to the capacitances in play, so it can't matter where you put the snubbers - they are on the equivalent of a bus-bar whether you put them distributed along or all at one end. It's not as if there can be a gradient of induced voltage along the cable!
With 8 x 0.1uF not completely solving the problem, that implies the impedance of the coupling between L and SL is at least equivalent, which I find hard to believe.
The only thing I can think of is that somehow this ghost voltage is partly or mainly dc, the capacitors only dealing with the ac bit of course.
Is there any way the timer could be responsible for a dc component?
I would like to suggest that, if possible, an insulation resistance test is done on this circuit.
And I think it would be useful to map out the path of permanent live and switched live, in terms of where those conductors run parallel to each other in the cables, and see if an alternative source of the power, or an alternative way of wiring, removing most of those parallel paths, could reduce the impedance of the ghost voltage, if that's what it is!