Markets are great for short term simple extrapolative prediction. At long term and/or non-continuous prediction they seem to have had significant failures.
Maybe this is optimism bias writ large? Another possibility is that even if there is robust demand despite certain destruction, these houses are really worth many millions for the short term (for cachet, status, etc.). Here we model them as really expensive vacations.
High P/E in the case of AMZN is not measure of long term. It's measure fo fast expected growth, low interest rates, shadow margins, and general short therm overconfidence.
It took 10 years for AMZN stock to recover from the peak of 1999 crash. There was nothing wrong with the company, markets just didn't want to wait.