I dunno if it was mentioned before but another, however limited, possibility is to rely on the RCS spikes. Every surface of stealth plane gives out full radar return (sans absorbed part) from a certain angle. One of mainstays of radar stealth design is allingment of various antennas/probes and other small objects that are integral part of a combat system and are protruding from the otherwise smooth fuselage to be aligned with bigger surfaces in such a way that they give a return from only few angles.
While in the past no care was payed to them as RCS given by the other parts of the plane was far bigger, once they became a proportionally big source of radar return all of the stealth planes (including f22 and f35) were designed in such a way that all of those small protruding elements are facing as few differnt sides as possible, ideally make them all face a single side and give a direct radar return when hit from only one, precise angle.
In practice, that means instead of getting many small spikes of radar return as the plane is flying around/over/alongside the radar thats monitoring it - that radar would get maybe one or two large spikes. While giving out a proportionally huge RCS at that one moment, overall such tactic is much better as you basically get just a momentary flash of a target, instead of series of glints as it goes along its path, making it more trackable.
BUT. If the mission of the stealth plane is a deep strike, where it has to go into the enemy radar network - say a b2 or f22 carrying jdams or SDBs - those spikes are a sure thing. It is virtually guaranteed that a radar will get at least one and probably two such spikes from one target if the target passes by it, over an arc of 120 or more degrees. Now that alone is still not very useful if you want to even begin to track planes. But, if you carefully position a network of radars over the area that attacker has to fly through, integrating them into a system where the computer knows where each radar is and is getting raw data from each radar, then it can use all that data, all those various spikes and compute a more accurate picture with far more points along plane's path where it was detected by its spike.
Granted, that's still only partial tracking, it won't give you a firing solution for radar guided SAMs, but it will offer better odds for your defence network as you'll be able to vector in your fighters to a right area to find a stealth plane. Alternatively, if you have enough spikes in a given time to get a really precise path, lets say update of course every 10-20 secs or so, you could even try launching a fair number of LOAL IIR missiles at a general area, hoping one or more would lock on the target and destroy it.
In the end, its far from a perfect solution but i believe it provides for a valid alternative for partial tracking that long wave radars give. Benefit is that one can use smaller, much more mobile radars for such network compare to long wave radars. Downside is price/manpower needed for a vast number of radars needed and maintaing such an integrated network.