Points 1-4 seem fine. Furthermore, I should point out that if you consider recent benchmarks, the bottleneck was not the DB but the framework on top (ie. Django or ROR). Generally DBs are extremely well designed.
Furthermore, it is still possible to cluster DBs as long as there is one write master and multiple read masters (of course, this does not apply when there is a high ratio of writes).
I am not an expert on clustering so you should check that last statement.
For statement #5: If your are concerned about SPs, I would strongly advise against MySQL. SPs in MySQL are very new and thus their implementation may not be quite up to par. You might want to consider PostgreSQL. The pgplsql SP language is supposed to be nearly equivalent to Oracle SPs. Furthermore, postgres also allows SPs in python, tcl, and perl which should cover pretty much all use cases for SPs though you sacrifice portability with languages other than pgplsql. Postgres also has much better scaling performance.
Furthermore, it is still possible to cluster DBs as long as there is one write master and multiple read masters (of course, this does not apply when there is a high ratio of writes).
I am not an expert on clustering so you should check that last statement.
For statement #5: If your are concerned about SPs, I would strongly advise against MySQL. SPs in MySQL are very new and thus their implementation may not be quite up to par. You might want to consider PostgreSQL. The pgplsql SP language is supposed to be nearly equivalent to Oracle SPs. Furthermore, postgres also allows SPs in python, tcl, and perl which should cover pretty much all use cases for SPs though you sacrifice portability with languages other than pgplsql. Postgres also has much better scaling performance.