All suggested keyword changes go into the &gen section...

If your wavefunction doesn't converge because:

(0) you started with a lousy initial geometry.
Start with a better geometry. Obviously.

(1) it ran out of iterations (default is 48) and the energy is still going down.
You can set maxit to a higher number (most of set it to 99 although sometimes you might need 199) in the restart (usually .01.in) file and then run the restart file.

(2) it ran out of iterations and the energy is hardly changing and going back and forth between two points.
You can loosen the convergence criteria: dconv and econv to a larger number (see the manual for the default numbers) to force the wavefunction to converge. After it succeeds, reset dconv and econv back in the restart file and run the restart file to convergence.

(3) there was a large energy change -- the most common problem!
There are many tricks to try here. All of them slow down your calculation somewhat so don't use them unless you have to. Try one first, then another, and if that still fails, a combination of them.
(i) Set iacscf=1. This temporarily shifts the virtual orbitals up in energy when the calculation starts and then gradually brings it back down as the calculation proceeds.
(ii) Set nops=1. Try to solve the wavefunction analytically instead of using the faster grid method.
(iii) Set vshift=1. Shifts the virtual orbitals up by a large amount. (Use in combination with nops=1 if the former fails.)
(iv) Set iconv=4. Use a difference convergence method. (There's also iconv=3 but it usually doesn't do better.)

(4) there was a large energy change and you have a system that has lots of unpaired electrons... [long answer]

(5) the geometry optimization died when Jaguar tries to change the Cartesian input into its own internal coordinates.
Set intopt=0 in the &gen section to prevent Jaguar from using internal coordinates.

(6) after many geometry optimization steps, the optimization gets "stuck".
Rerun the restart file (usually .01.in).