Having only one spawn location per side can be trouble; having two or more can take the edge off.
Clearcut: suffers from the above, but also not enough tactically significant terrain between (see: Sandblasted). If the central mountain -- more or less the one prominent position between the bases -- is taken, one side is firing into the other's base from cover, either pinning that side down in their base or forcing them to charge nearly a km without cover.
A change in method occurs to me: a string of capturable bases. These could either be organized in a "capture the last point to win" or in a circuit (which I don't think I've seen done before). Say, you could limit capturing to the front ~1-2 bases, and spawning to the front ~2-4, likely inhibiting spawn on a base that is being captured. (In the circuit model, a base at a point furthest from the front would flip to the side that loses one at the front.) Static defenses could be increased as one side is pushed further back. Perhaps a just-captured base could have restricted spawn/purchase/repair/rearm, weak to no static defense. (Maybe throw in some minor capturable bases providing a subset of radar, repair, rearm, light turret support, or whatever. Maybe some with little functionality but worth points, possibly more than you would expect from the lack of functionality, say a factory, dropship service pads -- some sort of longer-term strategic objectives. Some dedicated AI mech/tank defenders might add some pizazz to this, though maybe that's more than you want to get into -- but it would also allow for coop scenarios! Maybe something of a "sandbox" approach similar to the string-of-bases model(s), with ramping-but-randomized difficulty... Hey, I can dream.)
