Why $\lambda$‽: SIGMOD 2014

I totally failed to post a trip report on POPL 2014, and I'm now attending SIGMOD/PODS 2014, in Snowbird, Utah, so I'm writing the post during the conference to ensure I don't forget later. This is the first time I've been to SIGMOD in a while. There are usually 4-5 parallel tracks so it is impossible to see more than a fraction of the talks, but fortunately the talks I was most interested in were usually in separate sessions, so there were only a few tough decisions. There were many good talks, but several stood out:

Leonid Libkin's PODS keynote talk was on the gap between theory and practice for incomplete information. SQL uses a three-valued semantics for operations involving NULL values that has many well-explored problems. Database theory researchers have explored more principled approaches for incomplete information, mostly with high complexity. This talk (and paper) proposes the beginnings of a new approach that combines some previous attempts in new ways.
Eric Sedlar (Oracle Labs) gave the first SIGMOD invited talk, on "How I Learned to Stop Worrying and Love Compilers". Primarily about using compiler technology within databases to get higher-performance main memory database performance. The speaker also noted the continuing separation between PL and DB researchers.
Michael Benedikt presented work with ten Cate and Tsamoura on generating query reformulations (plans) from proofs. This approach is based on classical results by Craig (i.e. Craig interpolation) and appears to generalize several existing approaches.
Ioana Ileana presented her work with Cautis, Deutsch and Katsis on complete query reformulation using provenance-directed chase and backchase, with impressive experimental results. This seems related to Benedikt et al.'s paper, at least in rough outlines.
Maurice Herlihy gave the second SIGMOD invited talk, on "Fun with Transactional Memory". He gave an overview of transactional memory and how the idea of transactions was adapted from databases to concurrent programming and hardware, and then discussed several interesting and recent applications of hardware transactional memory capabilities (now in some shipping CPUs).
Jonathan Goldstein presented "Patience is a Virtue: Revisiting Merge and Sort on Modern Processors". Very interesting paper, examining an old, and underappreciated, approach to sorting that provides linear performance on sorted data and whose performance scales to n log n gradually as the input gets less sorted.
A panel of (primarily) industry people on "Are we experiencing a big data bubble?" The highlight (for me) was Daniel Abadi's presentation comparing NoSQL and related technologies with "instant gratification" and fast food.
Gabriel Bender presented work on "explainable security for relational databases". The motivation was app security and least privilege, but (being a SIGMOD paper) the paper actually considers explaining how to reformulate queries over security views, particularly identifying (hopefully) minimal additions to the privilege set that make a query answerable.
Sudeepa Roy presented work on "a formal approach to finding explanations for database queries" - actually about finding explanations for properties of query results. The basic idea is to identify deletions to the input which (once integrity constraints are restored) have a major impact on some predicate or value computed from the output. Interesting work and good presentation.