SQL Server is cost-based. That means Query Processor (QP) evaluates a number of scenarios to determine how it should attack the problem and how it should return the rows.
So even an insert/delete/update of 1 row can make a huge change between a scan or a seek in your execution plan.

SQL Server prefers scan as it per definition is sequential. Seeks are random in nature. The underlying disks (SSD or SCSI) doesn't matter. Yet.

You have probably encountered a tipping point where QP calcuates that a SEEK together with a KEY LOOKUP is less costly than a SCAN to get all columns in one sweep.