Chain Simulator tests fixes

[stefangutica]

Reasoning

• The chain simulator e2e tests failed intermittently, on every branch including development, and a different test failed each time.
• Duplicate websocket messages: under a slow handshake, the reconnect loop in the websocket test sent a second CONNECT on the same connection. The server opened a second socket for it, so every message arrived twice and the exact-count assertions failed.
• Transfers without action: the transaction pool warmer, enabled only in the e2e config, resolved every pending transaction and SCR each second. That included the ESDTTransfer SCR that delivers a newly issued token, which it looked up before the token was indexed. The miss was cached for an hour, so every later transfer of that token came back without action.
• Hidden errors and unstable order:
  • the transaction helpers swallowed errors and returned placeholders, so failures surfaced in unrelated assertions;
  • the test files not listed in the sequencer ran in whatever order jest discovered them.

Proposed Changes

• Websocket test:
  • retry a connection only after connect_error, subscribe once per client, and wait for the subscription acks;
  • count distinct messages only, since the broadcaster replays a round when it times out.
• E2E config: disabled transactionPoolWarmer, as on mainnet and devnet (/pool is still read on request), and enabled fastWarm.
• Issuance helpers: ESDT, NFT/SFT collection and Meta-ESDT issuance wait until the token is found through the same Elastic query the API runs (tokens/_search?q=_id:<id>). The Elastic URL is added to the test config (ELASTIC_URL).
• Test data: prepare-test-data waits on the API's own view (tokens, collections, NFTs, shards, node ratings) instead of a fixed 30s sleep.
• Error handling: sendTransaction, deploySc and getNonce rethrow instead of returning 'n/a' or 0, and deploySc fails when there is no SCDeploy event.
• Ordering and sleeps: unlisted test files run in alphabetical order, and the fixed sleep in the accounts NFT test is replaced by polling.

How to test

• Run the Chain simulator e2e tests workflow several times on this branch. A single green run proves little, since it passed intermittently before.
• Locally: npm run start-chain-simulator, docker compose up -d, npm run start:mainnet:e2e, npm run prepare:test-data, then npm run test:cs-e2e.

[github-actions]

k6 load testing comparison.
Base Commit Hash: 3d783b2
Target Commit Hash: edaf822

Metric	Base				Target				Diff
	Avg	Max	90	95	Avg	Max	90	95	Avg	Max	90	95
Tokens	40.51	153.81	57.35	60.07	28.20	1176.59	45.45	75.53	-30.39% ✅	+664.98% 🔴	-20.76% ✅	+25.75% 🔴
Nodes	62.78	15725.83	59.58	61.38	70.49	13505.25	66.32	87.30	+12.26% 🔴	-14.12% ✅	+11.31% 🔴	+42.22% 🔴
Mex	40.68	239.88	57.35	60.01	28.29	1729.92	44.19	73.96	-30.45% ✅	+621.17% 🔴	-22.95% ✅	+23.24% 🔴
Blocks	48.97	1698.19	59.27	61.42	37.60	1532.97	61.57	115.17	-23.21% ✅	-9.73% ✅	+3.88% 🔴	+87.51% 🔴
Accounts	41.76	928.05	57.83	60.16	29.10	2954.04	40.01	59.05	-30.32% ✅	+218.30% 🔴	-30.82% ✅	-1.85% ✅
Pool	40.93	632.56	57.60	60.06	28.14	707.60	45.73	77.50	-31.25% ✅	+11.86% 🔴	-20.61% ✅	+29.04% 🔴
Transactions	67.18	14408.73	59.91	61.50	52.18	8997.48	54.92	119.78	-22.33% ✅	-37.56% ✅	-8.33% ✅	+94.77% 🔴
Test Run Duration	60005.36				60003.15

Legend: Avg - Average Response Time, Max - Maximum Response Time, 90 - 90th Percentile, 95 - 95th Percentile
All times are in milliseconds.