Well, if we are going to be semi-realistic, then "purpose built" means "adapted from some SoC reference design".
Which leads us back to the numerous "which is the currently recommended device for OpenWrt" threads. AFICS, none of these have come to an absolute answer yet. There have never been a SoC with 100% pefect open source score. There have been, and still are a few nice, candidates. But my claim is that there will never be a perfect SoC as long as there are no SoC vendors actively participating in OpenWrt development. My proposal is that no hardware design is blessed with the "OpenWrt" name unless the SoC vendors BSP/SDK is identical to the current OpenWrt main branch.
Taking some OpenWrt snapshot and doing in-house development on it for years just doesn't cut it, as Qualcomm so nicely have demonstrated. There is still an infinite amount of work left before those SoCs can be considered supported in OpenWrt.
And depeloping gigantic drivers completly outside the Linux and OpenWrt communities leaves even more work to be done, as Mediatek still shows us.
And taking just one design and opening up some of the drivers, without any documentation of the vast number of firmware interfaces at all, will not magically create any community. Offering free hardware to developers makes absolutely no difference under such conditions, as Marvell have shown us.
Having support for old designs without following up with new designs doesn't do much good either, as Intel has shown.
There really isn't much to say about Broadcom, except it's a wonder they till exist. I have no idea why, but there isn't a chance in hell it's going to last so there is no point in considering them for any future OpenWrt work