If you don't mind all the AC on the line being presented to each driver, then transformers can be good.
The equivalent circuit might be, voltage sources driving transformer primary windings; secondaries connected in series, with the DC source on one side, and each DC load strung along.
All the DC loads need to be filtered, otherwise they load down the outputs.
Signal at the end is the sum of all oscillators.
If you need them filtered individually, consider a diplexing filter. You will need two for each node: one for DC (also known as a bias tee), and one for the AC signal. The DC filter has a high-pass low-pass characteristic, while the AC signal part will have a notch-bandpass characteristic.
Diplexing filters are designed as normal filters are (i.e., go find a filter calculator for e.g. Butterworth designs), but adjusted slightly: instead of both ends being matched impedance, one end is high impedance. The Hi-Z ends of multiple filters can be connected in parallel, and if the filters are complementary (i.e., one passes frequencies that the other one cuts), then the total response will be correct (when driven by a matched impedance). (At least, that was what I read. I don't know a proof of this method.)
For extracting DC from a combined AC and DC line, you simply pair high-pass and low-pass filters together. Or, since you don't need to split the AC at that point, even just the LPF is enough. Easy!
Tapping AC is going to be harder, because you need a diplexer designed for each oscillator frequency, and each diplexer needs to be different.
If you can tell us more about your application, we may be able to greatly simplify your circuit. Typically, signals are not combined at full power, but combined at a small signal level (where the attenuation of a resistor divider, for example, is easily compensated with an amplifier), then boosted to power output levels with a general purpose power amplifier.
Tim