Notes and ideas

Add inrush current limiting. Most USB ports / hubs should have protection already implemented in them but it certainly wouldn't hurt... and is not expensive. AP2337 seems like a good solution for IV-6 and IV-22 clocks and something along the lines of AP2411 for IV-12.
Add optional pulldown resistors for anodes. Some tubes are faulty so that some of the segments dimly light up that should be off. 100k resistor to the ground mitigates that issue at the expense of some extra current.
Fuse should be placed before ESD protection. There's very little chance (by the nature of it) for the ESD protection to fail short but it's a design flaw nonetheless. IV-12 version has this change already implemented.
Replacing 3 x 10uF booster output capacitors with 2 x 22uF capacitor.
PWM resolution and refresh rate is limited by ESP8266 capabilities... replacing ESP8266 with ESP32 would allow for even faster refresh rate and higher pwm resolution. Overall using SPI inside ISR function is not a particularly great idea (but it works and doesn't suffer from flicker caused by wifi) and could be avoided using ESP32. ESP8266 software Ticker can't get anywhere near that refresh rate (200uS per step)
implement soft start to smoothen inrush current.
Using smaller (SSOP-18) TD62783AFNG / TBD62783AFNG can allow for even smaller form factor.

IV-22 clock

Adding optional pulldown resistors for the segments - just like in the IV-12 version.

I would design it next time as a single-side design so everything would be reflown instead of havin to hand-solder the componetns at the bottom. The pcb would have to be larger but the design overall will be thinner which will make up for it. Using SSOP-18 version of the VFD driver makes alot of sense here.
Use larger booster inductor, good option is the one used on IV-12 clock

Despite the datasheet mentioning filament voltage to be DC, the digits are large enough to justify AC filament voltage. This can be done easily by adding TSC428 / MAX628 to the circuit and switching it via GPIO.