The first layer of a short TPU chain (about which, more later) came out vanishingly thin in the middle and much too thick on the ends:

So: let the platform cool, scrape off the wreckage, set the nozzle for Z=2.0 mm, and measure the actual gap at various spots across the platform.
Those results are the top set of measurements:

The bottom set of measurements came from a similar test a few days later, after pulling the BuildTak plate off, doing nothing other than scrutinizing it, reinstalling it, and successfully printing several TPU chains of varying design, none of which had any first-layer problems. The platform is slightly too high along the +Y and -Y edges (rear and front), with no bow worth mentioning.
My measurements are, perforce, done with a cold platform, for obvious reasons, and the TPU prints at 50 °C. I have the uneasy feeling the heater / BuildTak magnetic base can bow upward in the middle while it heats, then flatten out after a while at a stable temperature. The good news: it’s not permanently bent.
More study is needed, including thinwall boxes after letting the platform soak at 50 °C for varying times.
Spam comments get trashed, so don’t bother. Comment moderation may cause a delay.