They are aiming for the gaps that bigger vehicles don't fill in constellations. That seems rather dangerous: easily taken out by a number of competitors.
I think there's a lot of space here. There's really no major proprietary, secret sauce, or otherwise inaccessible technology that gives any launcher--SpaceX included--an unrecoverable advantage over competitors. As SpaceX has shown, its mostly about how a company is run.
Yeah, small launchers is a crowded space, and most of them are going to drop out. And yeah, the general concept of many small launches vs fewer big launches has yet to be proven viable. But generally, the way Chris and his team run Astra is pretty logical: Focus on technology and cost, not revenue. What remains to be seen is whether or not their logic can ultimately be realized as revenue.
And to be clear, the concept of being the "gap filler" or secondary launch provider for a constellation is actually pretty attractive. Two examples:
1. Currently, the concept for constellation redundancy and reliability is either to put 1) too many satellites in the constellation (= starlink) or 2) on orbit spares (either in plane or in parking orbits). The former is a problem because it requires more capital (for both sats and launch), the latter is a problem because there's still time require for a spare satellite to position into its proper orbital slot,
plus the clock on satellite lifetime more or less starts ticking once its launched (primarily because of the radiation environment).
Now imagine a scenario where a quick-turn launcher could--within days if not hours--direct inject a replacement satellite into the proper orbital slot. Astra (or whoever) has a bunch of sats for various constellations in a warehouse, or even already integrated onto rockets. Then they get the call, roll out the rocket, and launch The Thing. That concept significantly reduces the number of sats needed on orbit to provide full service
and minimizes service level impact of the satellite outage.
2. The physics of orbital precession is unavoidable. For a LEO constellation, depending on insertion and final altitudes, the hypothetical "single launch" for all the sats (think: Starship) would result in a year or more before the constellation actually provides as-designed coverage. Given that, again, the lifetime clock starts turning at launch, that's a massive hit to a constellation's revenue and amortization.
With more, smaller launchers...or even more intelligently, a mix of big and small launchers, a constellation can optimize/balance their launch costs against their in-service date. Because there's an inevitable time function to launching a bunch of sats (you can't do it all at once) you can do a few big launches up front and allow them to start precessing, then fill out the rest of the constellation with smaller, more targeted launches such that by the time the latter launches are done, the first big launches have completed precessing.
There's also a corollary use case of not having to launch spares into a plane (= enabling more service sats per launch thus decreasing average cost-to-orbit) and still being able to efficiently recover from a DOA sat.