Deploying, Voting, RCs

When you're making a big software change, there are a list of ways common in the industry to evaluate its impact and mitigate the chance of failure:

1. Code review
2. Have a QA organization run tests
3. Simulate an old set of inputs on the new software and evaluate the results
4. Run it in production in "read-only" mode and track how the new behavior would differ
5. Roll it out to a small set of production customers first before going to 100%
6. Deploy it but have a fall-back plan in place

I'd prefer a platform that did more of the latter half of the list, rather than going for broke. I'd also rather have one that didn't tell me everything was fine, and produced status updates when things were not fine.

Voting

Accordingly, I have burned all my tiny stash of RC by removing my upvotes for the entire top-18 list of witnesses except for @clayop. (I left @utopian-io and @pfunk and @drakos who have been close to top 20.) In the absence of other information, I'm trying to vote on anybody who is in the next 30 who has (a) deployed 20.3 as of this afternoon, and (b) has a reasonable price feed.

My tiny SP won't make much difference, but I object to the witnesses making the choice they did to roll out HF20, and their continuing receipt of witness rewards during the period when the platform was nonfunctional for 99% of its users.

Unfortunately, I ran out of RC before being able to cast 30 upvotes, which just shows that the pain is not yet over.

RC shortage

At 11:21 am Pacific 9/26/2018, this post failed because I needed 18,139,771,424 RC according to the API response and had 2,941,827,464 according to steemd.com's estimate.

@marissalingen's maximum mana is 9,990,517,596, so she could not post an article of this length at all.

At 1:40pm, steemd's estimate is 7,718,270,776 RC; this corresponds to a rate of about 34 million RC per minute. So this post's cost (now 2,0532,820,752) is about 604 minutes, or 10 hours 4 minutes.