Once something sticks to the wall, they are absolutely ruthless at getting it to product-market fit and commercial success, but interestingly enough, they stop there and all further development is merely iterative. They usually have some of the first decent and usuable products in any market, but tend to go for full commercial exploitation and pure maintenance mode from there.
Redshift used to be an absolute game changer for the Big Data market (I still remember their 1 TB for under 999$/month claim to this day), but Snowflake or BigQuery are just much better products this day and they never made the architectural shift from the client/server based architecture.
Prime Video is an absolute success story despite their awful interface, UX and crappy metadata, just because they got a grip on the hardware, smartly cross-sold it and bought / produced some good & free content.
I could go on and on, but it seems to be company core DNA to stop at 80/20. Whether that's a good thing or not, I'm still unsure.