The R/C servo control signal is a pulse train. There are two parameters that describe a pulse train: pulse width (PW) and pulse repetition interval (PRI). PW is what controls the servo position. It is critical, and must be stable. PRI is not nearly so critical, at least with R/C servos: they will generally be happy with a PRI anywhere from 20 to 50 ms, and they don't mind if it varies some. (You will also see PRF, for pulse repetition frequency, usually in radar systems. PRI 20-50 ms corresponds to PRF of 50 down to 20 Hz.)
Any kind of instability in the pulse width (PW) will cause the servo to "buzz", as it chases the control signal. Such instability can be triggered by the HUGE current draw of a servo in motion, if the power supply AT THE CONTROL CIRCUIT has insufficient surge capacity. (A standard servo in controlled motion can easily draw 250 mA. A "buzzing" servo can draw half-amp spikes. Yes, this is from experience.)
Instability in the PRI is generally not a problem for an R/C servo.
The very first thing I'd do is hang about a 250 uF capacitor directly across the servo power leads. The second thing I'd do is scope the control signal, triggering on the rising edge, and looking at the problem ranges to see if the timing on the falling edge is varying AT ALL.
Anyone planning on playing with servos should start by wiring up a 555 and a one-transistor inverter, generating the servo control signal entirely in hardware, and playing with this. Watching a servo sitting there, buzzing, watching your power supply ammeter going all over the place, registering half-amp spikes, is INSTRUCTIVE. Taming that beast is even more so. Note: This circuit will also demonstrate the servo's tolerance of varying PRI: the simplest 555 pulse generator for this purpose will vary PRI as it varies PW.