I see that - compared to AVR hw - millis() in loop() are not reliable enough to do reliable sequential control.
On AVR I do timekeeping and display update in loop() with sth like "if millis()%1000==0" but this does not trigger reliably on ESP even in an otherwise empty loop() ... on the other hand on ESP I have to include a delay(10) in loop() to have OneButton Lib working (@andrew melvin: see the example fail without delay in loop).
On AVR I even had to pack an delay in the above mentioned if branch to avoid occasional double triggering
Looks like the timing is not yet "compatible", probably have to include "Ticker" for now.
Again, thanks